Model profile · DeepSeek

deepseek-v4-pro

Unknown weightsmid · registry tag 2026 benchmark-derived

Thin verified coverage

Reads as thin verified coverage across the resolved source data. Capability fit excludes speed and cost.

Visible coverage: 54.1%
Verified coverage: 54.1%
Spread: 92.5%
Last verified: Oct 3, 2026

55%capability fit

textcodedocument2 aliases34 official source links

Open compare

Data version

Current snapshot.

Published Oct 3, 2026Model list checked4 stale sources105 claim warnings9 providers · 1334 tracked models

The registry snapshot and page stamp are shown so a stale deploy is visible at a glance. View live data status.

Source-linked scores by benchmark

Each row keeps the benchmark source, source type, raw metric, and percentile inside its fair comparison set. Capability summaries exclude archived categories, speed, and cost; those rows remain below.

Thin verified coverageThis model currently reads as thin verified coverage across the resolved source data.

Chat / text35 benchmarks63.7%

Intelligence Index

AA · Chat / text · Combined

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #64 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 36
Percentile: 88.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `intelligenceIndex`.

88.6% percentile inside its fair comparison set

36Raw benchmark value

AA-Omniscience accuracy

AA · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #49 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 49.1%
Percentile: 89.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `omniscienceAccuracy`.

89.4% percentile inside its fair comparison set

49.1%Raw benchmark value

AA-Omniscience non-hallucination

AA · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #420 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 5.2%
Percentile: 7.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `omniscienceNonHallucination`.

7.5% percentile inside its fair comparison set

5.2%Raw benchmark value

Input price

AA · Chat / text · Speed / cost

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #247 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: $1.3 /1M input tokens
Percentile: 30.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `price1mInputTokens`.

30.5% percentile inside its fair comparison set

$1.3 /1M input tokensRaw benchmark value

Output price

AA · Chat / text · Speed / cost

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #224 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: $4 /1M output tokens
Percentile: 37%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `price1mOutputTokens`.

37% percentile inside its fair comparison set

$4 /1M output tokensRaw benchmark value

Output Speed

AA · Chat / text · Speed / cost

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #105 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 106.6 tokens/s
Percentile: 60.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `medianOutputTokensPerSecond`.

60.3% percentile inside its fair comparison set

106.6 tokens/sRaw benchmark value

Time to first token

AA · Chat / text · Speed / cost

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #77 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 1.75s
Percentile: 71.2%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `medianTimeToFirstTokenSeconds`.

71.2% percentile inside its fair comparison set

1.75sRaw benchmark value

Time to first answer token

AA · Chat / text · Speed / cost

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #136 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 20.52s
Percentile: 48.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `medianTimeToFirstAnswerTokenSeconds`.

48.5% percentile inside its fair comparison set

20.52sRaw benchmark value

Text Arena

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #55

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,458
Percentile: 85.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: overall. Source rank: #62. Votes: 57478. Organization: deepseek. License: MIT.

85.6% percentile inside its fair comparison set

1,458Raw benchmark valueCI 1,454 - 1,462

Text Arena · Creative Writing

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #44

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,446
Percentile: 88.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: creative_writing. Source rank: #48. Votes: 9721. Organization: deepseek. License: MIT.

88.5% percentile inside its fair comparison set

1,446Raw benchmark valueCI 1,439 - 1,453

Text Arena · English

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #59

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,467
Percentile: 84.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: english. Source rank: #65. Votes: 25209. Organization: deepseek. License: MIT.

84.6% percentile inside its fair comparison set

1,467Raw benchmark valueCI 1,462 - 1,472

Text Arena · Exclude Ties

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #56

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,462
Percentile: 85.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: exclude_ties. Source rank: #63. Votes: 43538. Organization: deepseek. License: MIT.

85.4% percentile inside its fair comparison set

1,462Raw benchmark valueCI 1,456 - 1,467

Text Arena · Hard Prompts

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #60

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,480
Percentile: 84.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: hard_prompts. Source rank: #65. Votes: 38475. Organization: deepseek. License: MIT.

84.3% percentile inside its fair comparison set

1,480Raw benchmark valueCI 1,476 - 1,485

Text Arena · Hard Prompts English

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #63

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,485
Percentile: 83.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: hard_prompts_english. Source rank: #67. Votes: 17791. Organization: deepseek. License: MIT.

83.4% percentile inside its fair comparison set

1,485Raw benchmark valueCI 1,479 - 1,491

Text Arena · Instruction Following

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #58

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,452
Percentile: 84.8%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: instruction_following. Source rank: #62. Votes: 19985. Organization: deepseek. License: MIT.

84.8% percentile inside its fair comparison set

1,452Raw benchmark valueCI 1,447 - 1,458

Text Arena · Longer Query

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #54

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,472
Percentile: 85.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: longer_query. Source rank: #58. Votes: 25896. Organization: deepseek. License: MIT.

85.1% percentile inside its fair comparison set

1,472Raw benchmark valueCI 1,466 - 1,477

Text Arena · Multi Turn

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #44

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,475
Percentile: 88.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: multi_turn. Source rank: #48. Votes: 10357. Organization: deepseek. License: MIT.

88.5% percentile inside its fair comparison set

1,475Raw benchmark valueCI 1,468 - 1,482

Text Arena · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #45

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,451
Percentile: 88.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: overall. Source rank: #48. Votes: 57478. Organization: deepseek. License: MIT.

88.3% percentile inside its fair comparison set

1,451Raw benchmark valueCI 1,447 - 1,455

Text Arena · Creative Writing · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #35

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,446
Percentile: 90.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: creative_writing. Source rank: #37. Votes: 9721. Organization: deepseek. License: MIT.

90.9% percentile inside its fair comparison set

1,446Raw benchmark valueCI 1,438 - 1,453

Text Arena · English · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #44

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,459
Percentile: 88.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: english. Source rank: #46. Votes: 25209. Organization: deepseek. License: MIT.

88.6% percentile inside its fair comparison set

1,459Raw benchmark valueCI 1,454 - 1,464

Text Arena · Exclude Ties · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #45

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,450
Percentile: 88.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: exclude_ties. Source rank: #48. Votes: 43538. Organization: deepseek. License: MIT.

88.3% percentile inside its fair comparison set

1,450Raw benchmark valueCI 1,445 - 1,455

Text Arena · Hard Prompts · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #53

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,461
Percentile: 86.2%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: hard_prompts. Source rank: #56. Votes: 38475. Organization: deepseek. License: MIT.

86.2% percentile inside its fair comparison set

1,461Raw benchmark valueCI 1,457 - 1,466

Text Arena · Hard Prompts English · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #49

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,466
Percentile: 87.2%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: hard_prompts_english. Source rank: #51. Votes: 17791. Organization: deepseek. License: MIT.

87.2% percentile inside its fair comparison set

1,466Raw benchmark valueCI 1,460 - 1,472

Text Arena · Instruction Following · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #51

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,445
Percentile: 86.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: instruction_following. Source rank: #54. Votes: 19985. Organization: deepseek. License: MIT.

86.7% percentile inside its fair comparison set

1,445Raw benchmark valueCI 1,439 - 1,450

Text Arena · Longer Query · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #52

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,457
Percentile: 85.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: longer_query. Source rank: #55. Votes: 25896. Organization: deepseek. License: MIT.

85.6% percentile inside its fair comparison set

1,457Raw benchmark valueCI 1,452 - 1,463

Text Arena · Multi Turn · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #35

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,467
Percentile: 90.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: multi_turn. Source rank: #37. Votes: 10357. Organization: deepseek. License: MIT.

90.9% percentile inside its fair comparison set

1,467Raw benchmark valueCI 1,460 - 1,474

Instruction following

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #54

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 62.4%
Percentile: 18.5%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: IF. Tasks scored: 4.

18.5% percentile inside its fair comparison set

62.4%Raw benchmark value

Language

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #39

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 78.1%
Percentile: 41.5%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Language. Tasks scored: 3.

41.5% percentile inside its fair comparison set

78.1%Raw benchmark value

Paraphrase

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #59

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 56.2%
Percentile: 10.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: paraphrase. Category: IF.

10.8% percentile inside its fair comparison set

56.2%Raw benchmark value

Simplify

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #54

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 55.9%
Percentile: 18.5%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: simplify. Category: IF.

18.5% percentile inside its fair comparison set

55.9%Raw benchmark value

Story generation

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #48

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 67.8%
Percentile: 27.7%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: story_generation. Category: IF.

27.7% percentile inside its fair comparison set

67.8%Raw benchmark value

Summarize

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #35

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 69.5%
Percentile: 47.7%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: summarize. Category: IF.

47.7% percentile inside its fair comparison set

69.5%Raw benchmark value

Connections

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #35

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 98%
Percentile: 47.7%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: connections. Category: Language.

47.7% percentile inside its fair comparison set

98%Raw benchmark value

Plot unscrambling

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #47

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 56.4%
Percentile: 29.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: plot_unscrambling. Category: Language.

29.2% percentile inside its fair comparison set

56.4%Raw benchmark value

Typos

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #37

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 80%
Percentile: 46.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: typos. Category: Language.

46.2% percentile inside its fair comparison set

80%Raw benchmark value

Coding28 benchmarks45.7%

SciCode

AA · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #77 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 51%
Percentile: 61.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `scicode`.

61.6% percentile inside its fair comparison set

51%Raw benchmark value

Text Arena · Coding

AR · Coding · Human

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #72

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,500
Percentile: 80.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: coding. Source rank: #78. Votes: 17114. Organization: deepseek. License: MIT.

80.9% percentile inside its fair comparison set

1,500Raw benchmark valueCI 1,494 - 1,506

Text Arena · Coding · No Style Control

AR · Coding · Human

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #63

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,469
Percentile: 83.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: coding. Source rank: #67. Votes: 17114. Organization: deepseek. License: MIT.

83.3% percentile inside its fair comparison set

1,469Raw benchmark valueCI 1,463 - 1,475

Vibe Code Bench v1.1

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #44 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 49.9%
Percentile: 53.8%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: vibe-code; provider: DeepSeek.

53.8% percentile inside its fair comparison set

49.9%Raw benchmark valueCI 40.6% - 59.3%

WebDev Arena

AR · Coding · Human

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #65

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,446
Percentile: 49.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: overall. Source rank: #68. Votes: 14042. Organization: deepseek. License: MIT.

49.6% percentile inside its fair comparison set

1,446Raw benchmark valueCI 1,440 - 1,452

Code Arena · Webdev Html

AR · Coding · Human

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #69

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,434
Percentile: 45.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: webdev-html. Source rank: #74. Votes: 1503. Organization: deepseek. License: MIT.

45.6% percentile inside its fair comparison set

1,434Raw benchmark valueCI 1,418 - 1,450

Code Arena · Webdev React

AR · Coding · Human

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #64

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,440
Percentile: 44.2%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: webdev-react. Source rank: #67. Votes: 9680. Organization: deepseek. License: MIT.

44.2% percentile inside its fair comparison set

1,440Raw benchmark valueCI 1,433 - 1,447

Code Migration

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #40 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 26.2%
Percentile: 42.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: code-migration; provider: DeepSeek.

42.6% percentile inside its fair comparison set

26.2%Raw benchmark valueCI 18.3% - 34.1%

ProgramBench

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #50 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 0%
Percentile: 51.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: programbench; provider: DeepSeek.

51.7% percentile inside its fair comparison set

0%Raw benchmark valueCI 0% - 0%

Terminal-Bench 2.1

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #55 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 50.2%
Percentile: 22.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: terminal-bench-2-1; provider: DeepSeek.

22.4% percentile inside its fair comparison set

50.2%Raw benchmark valueCI 47.3% - 53.1%

SWE-bench Verified

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #34 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 77.4%
Percentile: 61%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: swebench; provider: DeepSeek.

61% percentile inside its fair comparison set

77.4%Raw benchmark valueCI 73.7% - 81.1%

LiveCodeBench

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #14 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 87.5%
Percentile: 89.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: lcb; provider: DeepSeek.

89.6% percentile inside its fair comparison set

87.5%Raw benchmark valueCI 85.6% - 89.4%

Agentic coding

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #57

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 42.6%
Percentile: 15.4%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Agentic Coding. Tasks scored: 3.

15.4% percentile inside its fair comparison set

42.6%Raw benchmark value

Coding

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #60

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 70%
Percentile: 9.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Coding. Tasks scored: 2.

9.2% percentile inside its fair comparison set

70%Raw benchmark value

JavaScript

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #60

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 54.5%
Percentile: 12.3%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: javascript. Category: Agentic Coding.

12.3% percentile inside its fair comparison set

54.5%Raw benchmark value

TypeScript

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #56

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 23.3%
Percentile: 18.5%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: typescript. Category: Agentic Coding.

18.5% percentile inside its fair comparison set

23.3%Raw benchmark value

Python

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #38

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 50%
Percentile: 46.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: python. Category: Agentic Coding.

46.2% percentile inside its fair comparison set

50%Raw benchmark value

Coding generation

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #59

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 70.4%
Percentile: 10.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: code_generation. Category: Coding.

10.8% percentile inside its fair comparison set

70.4%Raw benchmark value

Coding completion

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #56

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 69.6%
Percentile: 15.4%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: code_completion. Category: Coding.

15.4% percentile inside its fair comparison set

69.6%Raw benchmark value

Debugging

BB · Coding · Rubric

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #25 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 25.5%
Percentile: 0%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-debugging.

0% percentile inside its fair comparison set

25.5%Raw benchmark value

Refactoring

BB · Coding · Rubric

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #20 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 48.7%
Percentile: 9.5%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-refactoring.

9.5% percentile inside its fair comparison set

48.7%Raw benchmark value

Security

BB · Coding · Rubric

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #26 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 3.3%
Percentile: 0%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-security.

0% percentile inside its fair comparison set

3.3%Raw benchmark value

Speed throughput

BB · Coding · Speed / cost

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #19 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 87.9 t/s
Percentile: 28%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-throughput.

28% percentile inside its fair comparison set

87.9 t/sRaw benchmark value

Speed TTFT

BB · Coding · Speed / cost

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #20 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 6815.00ms
Percentile: 24%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-ttft.

24% percentile inside its fair comparison set

6815.00msRaw benchmark value

Speed cost

BB · Coding · Speed / cost

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #9 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: $0.1 $
Percentile: 68%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-speed-cost.

68% percentile inside its fair comparison set

$0.1 $Raw benchmark value

Overall quality

BB · Coding · Combined

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #22 · Source label: DeepSeek V4 Pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 16.3%
Percentile: 4.5%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-overall-quality.

4.5% percentile inside its fair comparison set

16.3%Raw benchmark value

Overall vibe

BB · Coding · Combined

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #23 · Source label: DeepSeek V4 Pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 20.1%
Percentile: 0%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-overall-vibe.

0% percentile inside its fair comparison set

20.1%Raw benchmark value

Overall coverage

BB · Coding · Combined

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #22 · Source label: DeepSeek V4 Pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 85.7%
Percentile: 22.7%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-overall-coverage.

22.7% percentile inside its fair comparison set

85.7%Raw benchmark value

Reasoning / math / science19 benchmarks64.5%

Humanity's Last Exam

AA · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #57 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 41%
Percentile: 89.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `hle`.

89.4% percentile inside its fair comparison set

41%Raw benchmark value

GPQA

AA · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #25 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 92.8%
Percentile: 95.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `gpqa`.

95.3% percentile inside its fair comparison set

92.8%Raw benchmark value

CritPt

AA · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #50 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 18%
Percentile: 89.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `critpt`.

89.5% percentile inside its fair comparison set

18%Raw benchmark value

Text Arena · Math

AR · Reasoning / math / science · Human

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #74

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,445
Percentile: 79.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: math. Source rank: #81. Votes: 3050. Organization: deepseek. License: MIT.

79.7% percentile inside its fair comparison set

1,445Raw benchmark valueCI 1,434 - 1,456

Text Arena · Math · No Style Control

AR · Reasoning / math / science · Human

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #68

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,443
Percentile: 81.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: math. Source rank: #72. Votes: 3050. Organization: deepseek. License: MIT.

81.3% percentile inside its fair comparison set

1,443Raw benchmark valueCI 1,432 - 1,454

ProofBench

VALS-AI · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #37 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 16%
Percentile: 14.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: proof_bench; provider: DeepSeek.

14.3% percentile inside its fair comparison set

16%Raw benchmark valueCI 8.8% - 23.2%

GPQA Diamond

VALS-AI · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #30 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 89.4%
Percentile: 76.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: gpqa; provider: DeepSeek.

76.6% percentile inside its fair comparison set

89.4%Raw benchmark valueCI 86.2% - 92.6%

MMLU Pro

VALS-AI · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #32 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 87.2%
Percentile: 75.2%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: mmlu_pro; provider: DeepSeek.

75.2% percentile inside its fair comparison set

87.2%Raw benchmark valueCI 86.6% - 87.9%

Mathematics

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #32

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 90.7%
Percentile: 52.3%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Mathematics. Tasks scored: 4.

52.3% percentile inside its fair comparison set

90.7%Raw benchmark value

Reasoning

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #45

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 82.7%
Percentile: 32.3%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Reasoning. Tasks scored: 4.

32.3% percentile inside its fair comparison set

82.7%Raw benchmark value

AMPS Hard

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #15

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 98%
Percentile: 83.1%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: AMPS_Hard. Category: Mathematics.

83.1% percentile inside its fair comparison set

98%Raw benchmark value

Integrals with game

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #34

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 78%
Percentile: 50.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: integrals_with_game. Category: Mathematics.

50.8% percentile inside its fair comparison set

78%Raw benchmark value

Math competition

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #23

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 96.1%
Percentile: 66.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: math_comp. Category: Mathematics.

66.2% percentile inside its fair comparison set

96.1%Raw benchmark value

Olympiad

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #33

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 90.6%
Percentile: 50.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: olympiad. Category: Mathematics.

50.8% percentile inside its fair comparison set

90.6%Raw benchmark value

Theory of mind

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #25

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 80.8%
Percentile: 66.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: theory_of_mind. Category: Reasoning.

66.2% percentile inside its fair comparison set

80.8%Raw benchmark value

Zebra puzzle

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #42

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 92%
Percentile: 36.9%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: zebra_puzzle. Category: Reasoning.

36.9% percentile inside its fair comparison set

92%Raw benchmark value

Spatial

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #55

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 94%
Percentile: 16.9%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: spatial. Category: Reasoning.

16.9% percentile inside its fair comparison set

94%Raw benchmark value

Logic with navigation

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #60

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 64%
Percentile: 10.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: logic_with_navigation. Category: Reasoning.

10.8% percentile inside its fair comparison set

64%Raw benchmark value

Reasoning

BB · Reasoning / math / science · Rubric

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #21 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 31.1%
Percentile: 4.8%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-reasoning.

4.8% percentile inside its fair comparison set

31.1%Raw benchmark value

Professional reasoning36 benchmarks36.4%

Text Arena · Expert

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena expert leaderboard.

Rank #70

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,480
Percentile: 78.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: expert. Source rank: #73. Votes: 5906. Organization: deepseek. License: MIT.

78.9% percentile inside its fair comparison set

1,480Raw benchmark valueCI 1,471 - 1,488

Text Arena · Industry Business And Management And Financial Operations

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_business_and_management_and_financial_operations leaderboard.

Rank #56

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,460
Percentile: 85.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_business_and_management_and_financial_operations. Source rank: #60. Votes: 11381. Organization: deepseek. License: MIT.

85.1% percentile inside its fair comparison set

1,460Raw benchmark valueCI 1,453 - 1,467

Text Arena · Industry Entertainment And Sports And Media

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_entertainment_and_sports_and_media leaderboard.

Rank #55

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,435
Percentile: 85.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_entertainment_and_sports_and_media. Source rank: #60. Votes: 12770. Organization: deepseek. License: MIT.

85.6% percentile inside its fair comparison set

1,435Raw benchmark valueCI 1,428 - 1,441

Text Arena · Industry Legal And Government

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_legal_and_government leaderboard.

Rank #46

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,474
Percentile: 87.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_legal_and_government. Source rank: #49. Votes: 4500. Organization: deepseek. License: MIT.

87.1% percentile inside its fair comparison set

1,474Raw benchmark valueCI 1,465 - 1,484

Text Arena · Industry Life And Physical And Social Science

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_life_and_physical_and_social_science leaderboard.

Rank #54

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,481
Percentile: 85.8%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_life_and_physical_and_social_science. Source rank: #56. Votes: 9142. Organization: deepseek. License: MIT.

85.8% percentile inside its fair comparison set

1,481Raw benchmark valueCI 1,474 - 1,488

Text Arena · Industry Mathematical

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_mathematical leaderboard.

Rank #64

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,460
Percentile: 82.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_mathematical. Source rank: #67. Votes: 3288. Organization: deepseek. License: MIT.

82.3% percentile inside its fair comparison set

1,460Raw benchmark valueCI 1,449 - 1,471

Text Arena · Industry Medicine And Healthcare

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_medicine_and_healthcare leaderboard.

Rank #57

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,478
Percentile: 83.8%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_medicine_and_healthcare. Source rank: #62. Votes: 4050. Organization: deepseek. License: MIT.

83.8% percentile inside its fair comparison set

1,478Raw benchmark valueCI 1,468 - 1,488

Text Arena · Industry Software And It Services

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_software_and_it_services leaderboard.

Rank #65

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,492
Percentile: 83%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_software_and_it_services. Source rank: #70. Votes: 23627. Organization: deepseek. License: MIT.

83% percentile inside its fair comparison set

1,492Raw benchmark valueCI 1,486 - 1,497

Text Arena · Industry Writing And Literature And Language

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_writing_and_literature_and_language leaderboard.

Rank #50

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,451
Percentile: 86.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_writing_and_literature_and_language. Source rank: #53. Votes: 14128. Organization: deepseek. License: MIT.

86.9% percentile inside its fair comparison set

1,451Raw benchmark valueCI 1,444 - 1,457

Text Arena · Expert · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena expert leaderboard.

Rank #62

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,463
Percentile: 81.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: expert. Source rank: #65. Votes: 5906. Organization: deepseek. License: MIT.

81.3% percentile inside its fair comparison set

1,463Raw benchmark valueCI 1,455 - 1,471

Text Arena · Industry Business And Management And Financial Operations · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_business_and_management_and_financial_operations leaderboard.

Rank #47

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,443
Percentile: 87.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_business_and_management_and_financial_operations. Source rank: #50. Votes: 11381. Organization: deepseek. License: MIT.

87.5% percentile inside its fair comparison set

1,443Raw benchmark valueCI 1,437 - 1,450

Text Arena · Industry Entertainment And Sports And Media · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_entertainment_and_sports_and_media leaderboard.

Rank #44

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,430
Percentile: 88.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_entertainment_and_sports_and_media. Source rank: #46. Votes: 12770. Organization: deepseek. License: MIT.

88.5% percentile inside its fair comparison set

1,430Raw benchmark valueCI 1,423 - 1,437

Text Arena · Industry Legal And Government · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_legal_and_government leaderboard.

Rank #43

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,464
Percentile: 88%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_legal_and_government. Source rank: #45. Votes: 4500. Organization: deepseek. License: MIT.

88% percentile inside its fair comparison set

1,464Raw benchmark valueCI 1,455 - 1,474

Text Arena · Industry Life And Physical And Social Science · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_life_and_physical_and_social_science leaderboard.

Rank #41

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,470
Percentile: 89.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_life_and_physical_and_social_science. Source rank: #43. Votes: 9142. Organization: deepseek. License: MIT.

89.3% percentile inside its fair comparison set

1,470Raw benchmark valueCI 1,463 - 1,477

Text Arena · Industry Mathematical · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_mathematical leaderboard.

Rank #62

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,455
Percentile: 82.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_mathematical. Source rank: #65. Votes: 3288. Organization: deepseek. License: MIT.

82.9% percentile inside its fair comparison set

1,455Raw benchmark valueCI 1,444 - 1,466

Text Arena · Industry Medicine And Healthcare · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_medicine_and_healthcare leaderboard.

Rank #43

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,465
Percentile: 87.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_medicine_and_healthcare. Source rank: #44. Votes: 4050. Organization: deepseek. License: MIT.

87.9% percentile inside its fair comparison set

1,465Raw benchmark valueCI 1,455 - 1,475

Text Arena · Industry Software And It Services · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_software_and_it_services leaderboard.

Rank #59

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,469
Percentile: 84.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_software_and_it_services. Source rank: #62. Votes: 23627. Organization: deepseek. License: MIT.

84.6% percentile inside its fair comparison set

1,469Raw benchmark valueCI 1,463 - 1,474

Text Arena · Industry Writing And Literature And Language · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_writing_and_literature_and_language leaderboard.

Rank #42

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,446
Percentile: 89.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_writing_and_literature_and_language. Source rank: #45. Votes: 14128. Organization: deepseek. License: MIT.

89.1% percentile inside its fair comparison set

1,446Raw benchmark valueCI 1,440 - 1,453

Vals Index

VALS-AI · Professional reasoning · Combined

Weighted model performance across economically relevant Vals tasks.

Rank #34 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 39
Percentile: 17.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: vals_index; provider: DeepSeek.

17.5% percentile inside its fair comparison set

39Raw benchmark valueCI 36 - 41

Legal Research Bench

VALS-AI · Professional reasoning · Objective

Applied legal research tasks.

Rank #47 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 23.1%
Percentile: 32.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: legal_research; provider: DeepSeek.

32.4% percentile inside its fair comparison set

23.1%Raw benchmark valueCI 17.3% - 28.8%

Harvey's Legal Agent Benchmark

VALS-AI · Professional reasoning · Objective

Completing legal work with documents, spreadsheets, presentations, and file-system tools.

Rank #33 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 3.8%
Percentile: 53.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: hlab; provider: DeepSeek.

53.6% percentile inside its fair comparison set

3.8%Raw benchmark valueCI 0.9% - 6.6%

LegalBench

VALS-AI · Professional reasoning · Objective

Academic legal reasoning tasks.

Rank #84 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 80.3%
Percentile: 38.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: legal_bench; provider: DeepSeek.

38.5% percentile inside its fair comparison set

80.3%Raw benchmark valueCI 79.4% - 81.2%

Finance Agent v2

VALS-AI · Professional reasoning · Objective

Core financial analyst tasks for agentic models.

Rank #49 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 44.1%
Percentile: 30.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: fabv2; provider: DeepSeek.

30.4% percentile inside its fair comparison set

44.1%Raw benchmark valueCI 42.8% - 45.4%

MedCode

VALS-AI · Professional reasoning · Objective

Medical billing support and coding tasks.

Rank #63 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 40.5%
Percentile: 34.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: medcode; provider: DeepSeek.

34.7% percentile inside its fair comparison set

40.5%Raw benchmark valueCI 36.3% - 44.6%

MedScribe

VALS-AI · Professional reasoning · Objective

Administrative documentation support for doctors.

Rank #72 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 75.1%
Percentile: 26%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: medscribe; provider: DeepSeek.

26% percentile inside its fair comparison set

75.1%Raw benchmark valueCI 71.2% - 79.1%

Public Benefits Bench

VALS-AI · Professional reasoning · Objective

Answering SNAP benefits questions across the public-benefits lifecycle.

Rank #22 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 62.9%
Percentile: 53.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: public-benefits-bench; provider: DeepSeek.

53.3% percentile inside its fair comparison set

62.9%Raw benchmark valueCI 60.5% - 65.4%

SkillsBench

VALS-AI · Professional reasoning · Objective

Applied professional skills tasks.

Rank #20 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 51.3%
Percentile: 42.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: skillsbench; provider: DeepSeek.

42.4% percentile inside its fair comparison set

51.3%Raw benchmark valueCI 42.3% - 60.3%

TaxEval v2

VALS-AI · Professional reasoning · Objective

Answer quality on tax questions and responses.

Rank #66 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 72.1%
Percentile: 49.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: tax_eval_v2; provider: DeepSeek.

49.6% percentile inside its fair comparison set

72.1%Raw benchmark valueCI 70.4% - 73.8%

Data analysis

LB · Professional reasoning · Objective

Structured data manipulation and table reasoning accuracy.

Rank #37

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 74.5%
Percentile: 44.6%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Data Analysis. Tasks scored: 3.

44.6% percentile inside its fair comparison set

74.5%Raw benchmark value

Overall

LB · Professional reasoning · Objective

Average objective performance across LiveBench's current public category mix.

Rank #53

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 71.6%
Percentile: 20%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category averages included: 7.

20% percentile inside its fair comparison set

71.6%Raw benchmark value

Consecutive events

LB · Professional reasoning · Objective

Objective consecutive events score in LiveBench.

Rank #42

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 75.4%
Percentile: 36.9%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: consecutive_events. Category: Data Analysis.

36.9% percentile inside its fair comparison set

75.4%Raw benchmark value

Table join

LB · Professional reasoning · Objective

Objective table join score in LiveBench.

Rank #33

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 48.2%
Percentile: 50.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: tablejoin. Category: Data Analysis.

50.8% percentile inside its fair comparison set

48.2%Raw benchmark value

Table reformat

LB · Professional reasoning · Objective

Objective table reformat score in LiveBench.

Rank #1

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 100%
Percentile: 100%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: tablereformat. Category: Data Analysis.

100% percentile inside its fair comparison set

100%Raw benchmark value

Public Benefits Bench v1

VALS-AI · Professional reasoning · Objective

Answering public-benefits questions across the benefits lifecycle.

Rank #7 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 57.6%
Percentile: 50%
Last updated: stale
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: public-benefits-bench-v1; provider: DeepSeek.

50% percentile inside its fair comparison set

57.6%Raw benchmark valueCI 55.1% - 60.1%

Hallucination

BB · Professional reasoning · Rubric

Accuracy and fabrication resistance under prompts that invite unsupported claims.

Rank #35 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 51.1%
Percentile: 0%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-hallucination.

0% percentile inside its fair comparison set

51.1%Raw benchmark value

BS pushback

BB · Professional reasoning · Rubric

Resistance to confidently accepting bogus assumptions in expert-style prompts.

Rank #19 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 36%
Percentile: 5.6%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-pushback.

5.6% percentile inside its fair comparison set

36%Raw benchmark value

Long context2 benchmarks71.8%

Long Context Reasoning

AA · Long context · Objective

It checks whether long-context claims survive contact with retrieval, memory, or long-document tasks.

Rank #49 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 80.3%
Percentile: 89.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `lcr`.

89.7% percentile inside its fair comparison set

80.3%Raw benchmark value

CorpFin v2

VALS-AI · Long context · Objective

It checks whether long-context claims survive contact with retrieval, memory, or long-document tasks.

Rank #55 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 61.4%
Percentile: 53.8%
Last updated: aging
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: corp_fin_v2; provider: DeepSeek.

53.8% percentile inside its fair comparison set

61.4%Raw benchmark valueCI 59.5% - 63.3%

Multilingual18 benchmarksn/a

Text Arena · Chinese

AR · Multilingual · Human

Observed user preference in Arena's Text Arena chinese leaderboard.

Rank #73

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,489
Percentile: 79.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: chinese. Source rank: #79. Votes: 3384. Organization: deepseek. License: MIT.

79.5% percentile inside its fair comparison set

1,489Raw benchmark valueCI 1,478 - 1,499

Text Arena · French

AR · Multilingual · Human

Observed user preference in Arena's Text Arena french leaderboard.

Rank #42

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,485
Percentile: 84%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: french. Source rank: #46. Votes: 2205. Organization: deepseek. License: MIT.

84% percentile inside its fair comparison set

1,485Raw benchmark valueCI 1,470 - 1,499

Text Arena · German

AR · Multilingual · Human

Observed user preference in Arena's Text Arena german leaderboard.

Rank #39

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,468
Percentile: 86%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: german. Source rank: #43. Votes: 999. Organization: deepseek. License: MIT.

86% percentile inside its fair comparison set

1,468Raw benchmark valueCI 1,449 - 1,488

Text Arena · Japanese

AR · Multilingual · Human

Observed user preference in Arena's Text Arena japanese leaderboard.

Rank #29

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,457
Percentile: 88.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: japanese. Source rank: #31. Votes: 715. Organization: deepseek. License: MIT.

88.5% percentile inside its fair comparison set

1,457Raw benchmark valueCI 1,433 - 1,481

Text Arena · Korean

AR · Multilingual · Human

Observed user preference in Arena's Text Arena korean leaderboard.

Rank #49

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,419
Percentile: 80.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: korean. Source rank: #56. Votes: 1106. Organization: deepseek. License: MIT.

80.3% percentile inside its fair comparison set

1,419Raw benchmark valueCI 1,400 - 1,438

Text Arena · Non English

AR · Multilingual · Human

Observed user preference in Arena's Text Arena non_english leaderboard.

Rank #57

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,445
Percentile: 85.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: non_english. Source rank: #63. Votes: 32269. Organization: deepseek. License: MIT.

85.1% percentile inside its fair comparison set

1,445Raw benchmark valueCI 1,440 - 1,450

Text Arena · Polish

AR · Multilingual · Human

Observed user preference in Arena's Text Arena polish leaderboard.

Rank #38

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,466
Percentile: 81.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: polish. Source rank: #45. Votes: 1123. Organization: deepseek. License: MIT.

81.4% percentile inside its fair comparison set

1,466Raw benchmark valueCI 1,448 - 1,484

Text Arena · Russian

AR · Multilingual · Human

Observed user preference in Arena's Text Arena russian leaderboard.

Rank #53

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,460
Percentile: 84.8%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: russian. Source rank: #59. Votes: 6313. Organization: deepseek. License: MIT.

84.8% percentile inside its fair comparison set

1,460Raw benchmark valueCI 1,452 - 1,469

Text Arena · Spanish

AR · Multilingual · Human

Observed user preference in Arena's Text Arena spanish leaderboard.

Rank #51

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,458
Percentile: 80.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: spanish. Source rank: #56. Votes: 1726. Organization: deepseek. License: MIT.

80.7% percentile inside its fair comparison set

1,458Raw benchmark valueCI 1,442 - 1,473

Text Arena · Chinese · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena chinese leaderboard.

Rank #69

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,481
Percentile: 80.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: chinese. Source rank: #72. Votes: 3384. Organization: deepseek. License: MIT.

80.7% percentile inside its fair comparison set

1,481Raw benchmark valueCI 1,470 - 1,492

Text Arena · French · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena french leaderboard.

Rank #38

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,472
Percentile: 85.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: french. Source rank: #40. Votes: 2205. Organization: deepseek. License: MIT.

85.6% percentile inside its fair comparison set

1,472Raw benchmark valueCI 1,457 - 1,486

Text Arena · German · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena german leaderboard.

Rank #39

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,457
Percentile: 86%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: german. Source rank: #43. Votes: 999. Organization: deepseek. License: MIT.

86% percentile inside its fair comparison set

1,457Raw benchmark valueCI 1,437 - 1,476

Text Arena · Japanese · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena japanese leaderboard.

Rank #29

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,447
Percentile: 88.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: japanese. Source rank: #32. Votes: 715. Organization: deepseek. License: MIT.

88.5% percentile inside its fair comparison set

1,447Raw benchmark valueCI 1,423 - 1,470

Text Arena · Korean · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena korean leaderboard.

Rank #47

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,411
Percentile: 81.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: korean. Source rank: #53. Votes: 1106. Organization: deepseek. License: MIT.

81.1% percentile inside its fair comparison set

1,411Raw benchmark valueCI 1,392 - 1,430

Text Arena · Non English · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena non_english leaderboard.

Rank #48

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,439
Percentile: 87.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: non_english. Source rank: #52. Votes: 32269. Organization: deepseek. License: MIT.

87.5% percentile inside its fair comparison set

1,439Raw benchmark valueCI 1,434 - 1,443

Text Arena · Polish · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena polish leaderboard.

Rank #30

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,462
Percentile: 85.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: polish. Source rank: #32. Votes: 1123. Organization: deepseek. License: MIT.

85.4% percentile inside its fair comparison set

1,462Raw benchmark valueCI 1,444 - 1,479

Text Arena · Russian · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena russian leaderboard.

Rank #41

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,453
Percentile: 88.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: russian. Source rank: #46. Votes: 6313. Organization: deepseek. License: MIT.

88.3% percentile inside its fair comparison set

1,453Raw benchmark valueCI 1,445 - 1,461

Text Arena · Spanish · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena spanish leaderboard.

Rank #37

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,458
Percentile: 86.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: spanish. Source rank: #38. Votes: 1726. Organization: deepseek. License: MIT.

86.1% percentile inside its fair comparison set

1,458Raw benchmark valueCI 1,442 - 1,473

Source links and registry checks

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Artificial Analysis

Oct 3, 2026

source →

official

LiveBench

Oct 3, 2026

source →

Model profile · DeepSeek

deepseek-v4-pro

Unknown weightsmid · registry tag 2026 benchmark-derived

Thin verified coverage

Reads as thin verified coverage across the resolved source data. Capability fit excludes speed and cost.

Visible coverage: 54.1%
Verified coverage: 54.1%
Spread: 92.5%
Last verified: Oct 3, 2026

55%capability fit

textcodedocument2 aliases34 official source links

Open compare

Data version

Current snapshot.

Published Oct 3, 2026Model list checked4 stale sources105 claim warnings9 providers · 1334 tracked models

The registry snapshot and page stamp are shown so a stale deploy is visible at a glance. View live data status.

Source-linked scores by benchmark

Thin verified coverageThis model currently reads as thin verified coverage across the resolved source data.

Chat / text35 benchmarks63.7%

Intelligence Index

AA · Chat / text · Combined

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #64 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 36
Percentile: 88.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `intelligenceIndex`.

88.6% percentile inside its fair comparison set

36Raw benchmark value

AA-Omniscience accuracy

AA · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #49 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 49.1%
Percentile: 89.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `omniscienceAccuracy`.

89.4% percentile inside its fair comparison set

49.1%Raw benchmark value

AA-Omniscience non-hallucination

AA · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #420 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 5.2%
Percentile: 7.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `omniscienceNonHallucination`.

7.5% percentile inside its fair comparison set

5.2%Raw benchmark value

Input price

AA · Chat / text · Speed / cost

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #247 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: $1.3 /1M input tokens
Percentile: 30.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `price1mInputTokens`.

30.5% percentile inside its fair comparison set

$1.3 /1M input tokensRaw benchmark value

Output price

AA · Chat / text · Speed / cost

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #224 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: $4 /1M output tokens
Percentile: 37%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `price1mOutputTokens`.

37% percentile inside its fair comparison set

$4 /1M output tokensRaw benchmark value

Output Speed

AA · Chat / text · Speed / cost

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #105 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 106.6 tokens/s
Percentile: 60.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `medianOutputTokensPerSecond`.

60.3% percentile inside its fair comparison set

106.6 tokens/sRaw benchmark value

Time to first token

AA · Chat / text · Speed / cost

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #77 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 1.75s
Percentile: 71.2%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `medianTimeToFirstTokenSeconds`.

71.2% percentile inside its fair comparison set

1.75sRaw benchmark value

Time to first answer token

AA · Chat / text · Speed / cost

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #136 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 20.52s
Percentile: 48.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `medianTimeToFirstAnswerTokenSeconds`.

48.5% percentile inside its fair comparison set

20.52sRaw benchmark value

Text Arena

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #55

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,458
Percentile: 85.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: overall. Source rank: #62. Votes: 57478. Organization: deepseek. License: MIT.

85.6% percentile inside its fair comparison set

1,458Raw benchmark valueCI 1,454 - 1,462

Text Arena · Creative Writing

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #44

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,446
Percentile: 88.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: creative_writing. Source rank: #48. Votes: 9721. Organization: deepseek. License: MIT.

88.5% percentile inside its fair comparison set

1,446Raw benchmark valueCI 1,439 - 1,453

Text Arena · English

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #59

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,467
Percentile: 84.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: english. Source rank: #65. Votes: 25209. Organization: deepseek. License: MIT.

84.6% percentile inside its fair comparison set

1,467Raw benchmark valueCI 1,462 - 1,472

Text Arena · Exclude Ties

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #56

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,462
Percentile: 85.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: exclude_ties. Source rank: #63. Votes: 43538. Organization: deepseek. License: MIT.

85.4% percentile inside its fair comparison set

1,462Raw benchmark valueCI 1,456 - 1,467

Text Arena · Hard Prompts

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #60

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,480
Percentile: 84.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: hard_prompts. Source rank: #65. Votes: 38475. Organization: deepseek. License: MIT.

84.3% percentile inside its fair comparison set

1,480Raw benchmark valueCI 1,476 - 1,485

Text Arena · Hard Prompts English

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #63

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,485
Percentile: 83.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: hard_prompts_english. Source rank: #67. Votes: 17791. Organization: deepseek. License: MIT.

83.4% percentile inside its fair comparison set

1,485Raw benchmark valueCI 1,479 - 1,491

Text Arena · Instruction Following

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #58

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,452
Percentile: 84.8%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: instruction_following. Source rank: #62. Votes: 19985. Organization: deepseek. License: MIT.

84.8% percentile inside its fair comparison set

1,452Raw benchmark valueCI 1,447 - 1,458

Text Arena · Longer Query

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #54

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,472
Percentile: 85.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: longer_query. Source rank: #58. Votes: 25896. Organization: deepseek. License: MIT.

85.1% percentile inside its fair comparison set

1,472Raw benchmark valueCI 1,466 - 1,477

Text Arena · Multi Turn

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #44

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,475
Percentile: 88.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: multi_turn. Source rank: #48. Votes: 10357. Organization: deepseek. License: MIT.

88.5% percentile inside its fair comparison set

1,475Raw benchmark valueCI 1,468 - 1,482

Text Arena · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #45

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,451
Percentile: 88.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: overall. Source rank: #48. Votes: 57478. Organization: deepseek. License: MIT.

88.3% percentile inside its fair comparison set

1,451Raw benchmark valueCI 1,447 - 1,455

Text Arena · Creative Writing · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #35

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,446
Percentile: 90.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: creative_writing. Source rank: #37. Votes: 9721. Organization: deepseek. License: MIT.

90.9% percentile inside its fair comparison set

1,446Raw benchmark valueCI 1,438 - 1,453

Text Arena · English · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #44

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,459
Percentile: 88.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: english. Source rank: #46. Votes: 25209. Organization: deepseek. License: MIT.

88.6% percentile inside its fair comparison set

1,459Raw benchmark valueCI 1,454 - 1,464

Text Arena · Exclude Ties · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #45

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,450
Percentile: 88.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: exclude_ties. Source rank: #48. Votes: 43538. Organization: deepseek. License: MIT.

88.3% percentile inside its fair comparison set

1,450Raw benchmark valueCI 1,445 - 1,455

Text Arena · Hard Prompts · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #53

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,461
Percentile: 86.2%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: hard_prompts. Source rank: #56. Votes: 38475. Organization: deepseek. License: MIT.

86.2% percentile inside its fair comparison set

1,461Raw benchmark valueCI 1,457 - 1,466

Text Arena · Hard Prompts English · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #49

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,466
Percentile: 87.2%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: hard_prompts_english. Source rank: #51. Votes: 17791. Organization: deepseek. License: MIT.

87.2% percentile inside its fair comparison set

1,466Raw benchmark valueCI 1,460 - 1,472

Text Arena · Instruction Following · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #51

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,445
Percentile: 86.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: instruction_following. Source rank: #54. Votes: 19985. Organization: deepseek. License: MIT.

86.7% percentile inside its fair comparison set

1,445Raw benchmark valueCI 1,439 - 1,450

Text Arena · Longer Query · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #52

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,457
Percentile: 85.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: longer_query. Source rank: #55. Votes: 25896. Organization: deepseek. License: MIT.

85.6% percentile inside its fair comparison set

1,457Raw benchmark valueCI 1,452 - 1,463

Text Arena · Multi Turn · No Style Control

AR · Chat / text · Human

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #35

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,467
Percentile: 90.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: multi_turn. Source rank: #37. Votes: 10357. Organization: deepseek. License: MIT.

90.9% percentile inside its fair comparison set

1,467Raw benchmark valueCI 1,460 - 1,474

Instruction following

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #54

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 62.4%
Percentile: 18.5%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: IF. Tasks scored: 4.

18.5% percentile inside its fair comparison set

62.4%Raw benchmark value

Language

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #39

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 78.1%
Percentile: 41.5%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Language. Tasks scored: 3.

41.5% percentile inside its fair comparison set

78.1%Raw benchmark value

Paraphrase

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #59

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 56.2%
Percentile: 10.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: paraphrase. Category: IF.

10.8% percentile inside its fair comparison set

56.2%Raw benchmark value

Simplify

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #54

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 55.9%
Percentile: 18.5%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: simplify. Category: IF.

18.5% percentile inside its fair comparison set

55.9%Raw benchmark value

Story generation

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #48

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 67.8%
Percentile: 27.7%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: story_generation. Category: IF.

27.7% percentile inside its fair comparison set

67.8%Raw benchmark value

Summarize

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #35

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 69.5%
Percentile: 47.7%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: summarize. Category: IF.

47.7% percentile inside its fair comparison set

69.5%Raw benchmark value

Connections

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #35

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 98%
Percentile: 47.7%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: connections. Category: Language.

47.7% percentile inside its fair comparison set

98%Raw benchmark value

Plot unscrambling

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #47

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 56.4%
Percentile: 29.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: plot_unscrambling. Category: Language.

29.2% percentile inside its fair comparison set

56.4%Raw benchmark value

Typos

LB · Chat / text · Objective

It tests whether the model is actually useful in normal conversational turns, not just on narrow correctness tasks.

Rank #37

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 80%
Percentile: 46.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: typos. Category: Language.

46.2% percentile inside its fair comparison set

80%Raw benchmark value

Coding28 benchmarks45.7%

SciCode

AA · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #77 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 51%
Percentile: 61.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `scicode`.

61.6% percentile inside its fair comparison set

51%Raw benchmark value

Text Arena · Coding

AR · Coding · Human

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #72

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,500
Percentile: 80.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: coding. Source rank: #78. Votes: 17114. Organization: deepseek. License: MIT.

80.9% percentile inside its fair comparison set

1,500Raw benchmark valueCI 1,494 - 1,506

Text Arena · Coding · No Style Control

AR · Coding · Human

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #63

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,469
Percentile: 83.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: coding. Source rank: #67. Votes: 17114. Organization: deepseek. License: MIT.

83.3% percentile inside its fair comparison set

1,469Raw benchmark valueCI 1,463 - 1,475

Vibe Code Bench v1.1

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #44 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 49.9%
Percentile: 53.8%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: vibe-code; provider: DeepSeek.

53.8% percentile inside its fair comparison set

49.9%Raw benchmark valueCI 40.6% - 59.3%

WebDev Arena

AR · Coding · Human

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #65

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,446
Percentile: 49.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: overall. Source rank: #68. Votes: 14042. Organization: deepseek. License: MIT.

49.6% percentile inside its fair comparison set

1,446Raw benchmark valueCI 1,440 - 1,452

Code Arena · Webdev Html

AR · Coding · Human

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #69

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,434
Percentile: 45.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: webdev-html. Source rank: #74. Votes: 1503. Organization: deepseek. License: MIT.

45.6% percentile inside its fair comparison set

1,434Raw benchmark valueCI 1,418 - 1,450

Code Arena · Webdev React

AR · Coding · Human

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #64

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,440
Percentile: 44.2%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: webdev-react. Source rank: #67. Votes: 9680. Organization: deepseek. License: MIT.

44.2% percentile inside its fair comparison set

1,440Raw benchmark valueCI 1,433 - 1,447

Code Migration

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #40 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 26.2%
Percentile: 42.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: code-migration; provider: DeepSeek.

42.6% percentile inside its fair comparison set

26.2%Raw benchmark valueCI 18.3% - 34.1%

ProgramBench

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #50 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 0%
Percentile: 51.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: programbench; provider: DeepSeek.

51.7% percentile inside its fair comparison set

0%Raw benchmark valueCI 0% - 0%

Terminal-Bench 2.1

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #55 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 50.2%
Percentile: 22.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: terminal-bench-2-1; provider: DeepSeek.

22.4% percentile inside its fair comparison set

50.2%Raw benchmark valueCI 47.3% - 53.1%

SWE-bench Verified

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #34 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 77.4%
Percentile: 61%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: swebench; provider: DeepSeek.

61% percentile inside its fair comparison set

77.4%Raw benchmark valueCI 73.7% - 81.1%

LiveCodeBench

VALS-AI · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #14 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 87.5%
Percentile: 89.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: lcb; provider: DeepSeek.

89.6% percentile inside its fair comparison set

87.5%Raw benchmark valueCI 85.6% - 89.4%

Agentic coding

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #57

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 42.6%
Percentile: 15.4%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Agentic Coding. Tasks scored: 3.

15.4% percentile inside its fair comparison set

42.6%Raw benchmark value

Coding

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #60

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 70%
Percentile: 9.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Coding. Tasks scored: 2.

9.2% percentile inside its fair comparison set

70%Raw benchmark value

JavaScript

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #60

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 54.5%
Percentile: 12.3%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: javascript. Category: Agentic Coding.

12.3% percentile inside its fair comparison set

54.5%Raw benchmark value

TypeScript

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #56

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 23.3%
Percentile: 18.5%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: typescript. Category: Agentic Coding.

18.5% percentile inside its fair comparison set

23.3%Raw benchmark value

Python

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #38

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 50%
Percentile: 46.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: python. Category: Agentic Coding.

46.2% percentile inside its fair comparison set

50%Raw benchmark value

Coding generation

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #59

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 70.4%
Percentile: 10.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: code_generation. Category: Coding.

10.8% percentile inside its fair comparison set

70.4%Raw benchmark value

Coding completion

LB · Coding · Objective

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #56

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 69.6%
Percentile: 15.4%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: code_completion. Category: Coding.

15.4% percentile inside its fair comparison set

69.6%Raw benchmark value

Debugging

BB · Coding · Rubric

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #25 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 25.5%
Percentile: 0%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-debugging.

0% percentile inside its fair comparison set

25.5%Raw benchmark value

Refactoring

BB · Coding · Rubric

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #20 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 48.7%
Percentile: 9.5%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-refactoring.

9.5% percentile inside its fair comparison set

48.7%Raw benchmark value

Security

BB · Coding · Rubric

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #26 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 3.3%
Percentile: 0%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-security.

0% percentile inside its fair comparison set

3.3%Raw benchmark value

Speed throughput

BB · Coding · Speed / cost

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #19 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 87.9 t/s
Percentile: 28%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-throughput.

28% percentile inside its fair comparison set

87.9 t/sRaw benchmark value

Speed TTFT

BB · Coding · Speed / cost

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #20 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 6815.00ms
Percentile: 24%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-ttft.

24% percentile inside its fair comparison set

6815.00msRaw benchmark value

Speed cost

BB · Coding · Speed / cost

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #9 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: $0.1 $
Percentile: 68%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-speed-cost.

68% percentile inside its fair comparison set

$0.1 $Raw benchmark value

Overall quality

BB · Coding · Combined

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #22 · Source label: DeepSeek V4 Pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 16.3%
Percentile: 4.5%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-overall-quality.

4.5% percentile inside its fair comparison set

16.3%Raw benchmark value

Overall vibe

BB · Coding · Combined

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #23 · Source label: DeepSeek V4 Pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 20.1%
Percentile: 0%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-overall-vibe.

0% percentile inside its fair comparison set

20.1%Raw benchmark value

Overall coverage

BB · Coding · Combined

It tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.

Rank #22 · Source label: DeepSeek V4 Pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 85.7%
Percentile: 22.7%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-overall-coverage.

22.7% percentile inside its fair comparison set

85.7%Raw benchmark value

Reasoning / math / science19 benchmarks64.5%

Humanity's Last Exam

AA · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #57 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 41%
Percentile: 89.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `hle`.

89.4% percentile inside its fair comparison set

41%Raw benchmark value

GPQA

AA · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #25 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 92.8%
Percentile: 95.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `gpqa`.

95.3% percentile inside its fair comparison set

92.8%Raw benchmark value

CritPt

AA · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #50 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 18%
Percentile: 89.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `critpt`.

89.5% percentile inside its fair comparison set

18%Raw benchmark value

Text Arena · Math

AR · Reasoning / math / science · Human

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #74

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,445
Percentile: 79.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: math. Source rank: #81. Votes: 3050. Organization: deepseek. License: MIT.

79.7% percentile inside its fair comparison set

1,445Raw benchmark valueCI 1,434 - 1,456

Text Arena · Math · No Style Control

AR · Reasoning / math / science · Human

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #68

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,443
Percentile: 81.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: math. Source rank: #72. Votes: 3050. Organization: deepseek. License: MIT.

81.3% percentile inside its fair comparison set

1,443Raw benchmark valueCI 1,432 - 1,454

ProofBench

VALS-AI · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #37 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 16%
Percentile: 14.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: proof_bench; provider: DeepSeek.

14.3% percentile inside its fair comparison set

16%Raw benchmark valueCI 8.8% - 23.2%

GPQA Diamond

VALS-AI · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #30 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 89.4%
Percentile: 76.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: gpqa; provider: DeepSeek.

76.6% percentile inside its fair comparison set

89.4%Raw benchmark valueCI 86.2% - 92.6%

MMLU Pro

VALS-AI · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #32 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 87.2%
Percentile: 75.2%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: mmlu_pro; provider: DeepSeek.

75.2% percentile inside its fair comparison set

87.2%Raw benchmark valueCI 86.6% - 87.9%

Mathematics

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #32

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 90.7%
Percentile: 52.3%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Mathematics. Tasks scored: 4.

52.3% percentile inside its fair comparison set

90.7%Raw benchmark value

Reasoning

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #45

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 82.7%
Percentile: 32.3%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Reasoning. Tasks scored: 4.

32.3% percentile inside its fair comparison set

82.7%Raw benchmark value

AMPS Hard

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #15

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 98%
Percentile: 83.1%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: AMPS_Hard. Category: Mathematics.

83.1% percentile inside its fair comparison set

98%Raw benchmark value

Integrals with game

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #34

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 78%
Percentile: 50.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: integrals_with_game. Category: Mathematics.

50.8% percentile inside its fair comparison set

78%Raw benchmark value

Math competition

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #23

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 96.1%
Percentile: 66.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: math_comp. Category: Mathematics.

66.2% percentile inside its fair comparison set

96.1%Raw benchmark value

Olympiad

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #33

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 90.6%
Percentile: 50.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: olympiad. Category: Mathematics.

50.8% percentile inside its fair comparison set

90.6%Raw benchmark value

Theory of mind

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #25

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 80.8%
Percentile: 66.2%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: theory_of_mind. Category: Reasoning.

66.2% percentile inside its fair comparison set

80.8%Raw benchmark value

Zebra puzzle

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #42

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 92%
Percentile: 36.9%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: zebra_puzzle. Category: Reasoning.

36.9% percentile inside its fair comparison set

92%Raw benchmark value

Spatial

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #55

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 94%
Percentile: 16.9%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: spatial. Category: Reasoning.

16.9% percentile inside its fair comparison set

94%Raw benchmark value

Logic with navigation

LB · Reasoning / math / science · Objective

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #60

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 64%
Percentile: 10.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: logic_with_navigation. Category: Reasoning.

10.8% percentile inside its fair comparison set

64%Raw benchmark value

Reasoning

BB · Reasoning / math / science · Rubric

It is one of the cleaner reads on deliberate reasoning strength rather than style or popularity.

Rank #21 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 31.1%
Percentile: 4.8%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-reasoning.

4.8% percentile inside its fair comparison set

31.1%Raw benchmark value

Professional reasoning36 benchmarks36.4%

Text Arena · Expert

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena expert leaderboard.

Rank #70

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,480
Percentile: 78.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: expert. Source rank: #73. Votes: 5906. Organization: deepseek. License: MIT.

78.9% percentile inside its fair comparison set

1,480Raw benchmark valueCI 1,471 - 1,488

Text Arena · Industry Business And Management And Financial Operations

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_business_and_management_and_financial_operations leaderboard.

Rank #56

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,460
Percentile: 85.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_business_and_management_and_financial_operations. Source rank: #60. Votes: 11381. Organization: deepseek. License: MIT.

85.1% percentile inside its fair comparison set

1,460Raw benchmark valueCI 1,453 - 1,467

Text Arena · Industry Entertainment And Sports And Media

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_entertainment_and_sports_and_media leaderboard.

Rank #55

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,435
Percentile: 85.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_entertainment_and_sports_and_media. Source rank: #60. Votes: 12770. Organization: deepseek. License: MIT.

85.6% percentile inside its fair comparison set

1,435Raw benchmark valueCI 1,428 - 1,441

Text Arena · Industry Legal And Government

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_legal_and_government leaderboard.

Rank #46

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,474
Percentile: 87.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_legal_and_government. Source rank: #49. Votes: 4500. Organization: deepseek. License: MIT.

87.1% percentile inside its fair comparison set

1,474Raw benchmark valueCI 1,465 - 1,484

Text Arena · Industry Life And Physical And Social Science

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_life_and_physical_and_social_science leaderboard.

Rank #54

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,481
Percentile: 85.8%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_life_and_physical_and_social_science. Source rank: #56. Votes: 9142. Organization: deepseek. License: MIT.

85.8% percentile inside its fair comparison set

1,481Raw benchmark valueCI 1,474 - 1,488

Text Arena · Industry Mathematical

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_mathematical leaderboard.

Rank #64

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,460
Percentile: 82.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_mathematical. Source rank: #67. Votes: 3288. Organization: deepseek. License: MIT.

82.3% percentile inside its fair comparison set

1,460Raw benchmark valueCI 1,449 - 1,471

Text Arena · Industry Medicine And Healthcare

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_medicine_and_healthcare leaderboard.

Rank #57

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,478
Percentile: 83.8%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_medicine_and_healthcare. Source rank: #62. Votes: 4050. Organization: deepseek. License: MIT.

83.8% percentile inside its fair comparison set

1,478Raw benchmark valueCI 1,468 - 1,488

Text Arena · Industry Software And It Services

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_software_and_it_services leaderboard.

Rank #65

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,492
Percentile: 83%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_software_and_it_services. Source rank: #70. Votes: 23627. Organization: deepseek. License: MIT.

83% percentile inside its fair comparison set

1,492Raw benchmark valueCI 1,486 - 1,497

Text Arena · Industry Writing And Literature And Language

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_writing_and_literature_and_language leaderboard.

Rank #50

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,451
Percentile: 86.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_writing_and_literature_and_language. Source rank: #53. Votes: 14128. Organization: deepseek. License: MIT.

86.9% percentile inside its fair comparison set

1,451Raw benchmark valueCI 1,444 - 1,457

Text Arena · Expert · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena expert leaderboard.

Rank #62

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,463
Percentile: 81.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: expert. Source rank: #65. Votes: 5906. Organization: deepseek. License: MIT.

81.3% percentile inside its fair comparison set

1,463Raw benchmark valueCI 1,455 - 1,471

Text Arena · Industry Business And Management And Financial Operations · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_business_and_management_and_financial_operations leaderboard.

Rank #47

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,443
Percentile: 87.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_business_and_management_and_financial_operations. Source rank: #50. Votes: 11381. Organization: deepseek. License: MIT.

87.5% percentile inside its fair comparison set

1,443Raw benchmark valueCI 1,437 - 1,450

Text Arena · Industry Entertainment And Sports And Media · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_entertainment_and_sports_and_media leaderboard.

Rank #44

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,430
Percentile: 88.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_entertainment_and_sports_and_media. Source rank: #46. Votes: 12770. Organization: deepseek. License: MIT.

88.5% percentile inside its fair comparison set

1,430Raw benchmark valueCI 1,423 - 1,437

Text Arena · Industry Legal And Government · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_legal_and_government leaderboard.

Rank #43

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,464
Percentile: 88%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_legal_and_government. Source rank: #45. Votes: 4500. Organization: deepseek. License: MIT.

88% percentile inside its fair comparison set

1,464Raw benchmark valueCI 1,455 - 1,474

Text Arena · Industry Life And Physical And Social Science · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_life_and_physical_and_social_science leaderboard.

Rank #41

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,470
Percentile: 89.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_life_and_physical_and_social_science. Source rank: #43. Votes: 9142. Organization: deepseek. License: MIT.

89.3% percentile inside its fair comparison set

1,470Raw benchmark valueCI 1,463 - 1,477

Text Arena · Industry Mathematical · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_mathematical leaderboard.

Rank #62

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,455
Percentile: 82.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_mathematical. Source rank: #65. Votes: 3288. Organization: deepseek. License: MIT.

82.9% percentile inside its fair comparison set

1,455Raw benchmark valueCI 1,444 - 1,466

Text Arena · Industry Medicine And Healthcare · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_medicine_and_healthcare leaderboard.

Rank #43

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,465
Percentile: 87.9%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_medicine_and_healthcare. Source rank: #44. Votes: 4050. Organization: deepseek. License: MIT.

87.9% percentile inside its fair comparison set

1,465Raw benchmark valueCI 1,455 - 1,475

Text Arena · Industry Software And It Services · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_software_and_it_services leaderboard.

Rank #59

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,469
Percentile: 84.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_software_and_it_services. Source rank: #62. Votes: 23627. Organization: deepseek. License: MIT.

84.6% percentile inside its fair comparison set

1,469Raw benchmark valueCI 1,463 - 1,474

Text Arena · Industry Writing And Literature And Language · No Style Control

AR · Professional reasoning · Human

Observed user preference in Arena's Text Arena industry_writing_and_literature_and_language leaderboard.

Rank #42

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,446
Percentile: 89.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: industry_writing_and_literature_and_language. Source rank: #45. Votes: 14128. Organization: deepseek. License: MIT.

89.1% percentile inside its fair comparison set

1,446Raw benchmark valueCI 1,440 - 1,453

Vals Index

VALS-AI · Professional reasoning · Combined

Weighted model performance across economically relevant Vals tasks.

Rank #34 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 39
Percentile: 17.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: vals_index; provider: DeepSeek.

17.5% percentile inside its fair comparison set

39Raw benchmark valueCI 36 - 41

Legal Research Bench

VALS-AI · Professional reasoning · Objective

Applied legal research tasks.

Rank #47 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 23.1%
Percentile: 32.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: legal_research; provider: DeepSeek.

32.4% percentile inside its fair comparison set

23.1%Raw benchmark valueCI 17.3% - 28.8%

Harvey's Legal Agent Benchmark

VALS-AI · Professional reasoning · Objective

Completing legal work with documents, spreadsheets, presentations, and file-system tools.

Rank #33 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 3.8%
Percentile: 53.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: hlab; provider: DeepSeek.

53.6% percentile inside its fair comparison set

3.8%Raw benchmark valueCI 0.9% - 6.6%

LegalBench

VALS-AI · Professional reasoning · Objective

Academic legal reasoning tasks.

Rank #84 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 80.3%
Percentile: 38.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: legal_bench; provider: DeepSeek.

38.5% percentile inside its fair comparison set

80.3%Raw benchmark valueCI 79.4% - 81.2%

Finance Agent v2

VALS-AI · Professional reasoning · Objective

Core financial analyst tasks for agentic models.

Rank #49 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 44.1%
Percentile: 30.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: fabv2; provider: DeepSeek.

30.4% percentile inside its fair comparison set

44.1%Raw benchmark valueCI 42.8% - 45.4%

MedCode

VALS-AI · Professional reasoning · Objective

Medical billing support and coding tasks.

Rank #63 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 40.5%
Percentile: 34.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: medcode; provider: DeepSeek.

34.7% percentile inside its fair comparison set

40.5%Raw benchmark valueCI 36.3% - 44.6%

MedScribe

VALS-AI · Professional reasoning · Objective

Administrative documentation support for doctors.

Rank #72 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 75.1%
Percentile: 26%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: medscribe; provider: DeepSeek.

26% percentile inside its fair comparison set

75.1%Raw benchmark valueCI 71.2% - 79.1%

Public Benefits Bench

VALS-AI · Professional reasoning · Objective

Answering SNAP benefits questions across the public-benefits lifecycle.

Rank #22 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 62.9%
Percentile: 53.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: public-benefits-bench; provider: DeepSeek.

53.3% percentile inside its fair comparison set

62.9%Raw benchmark valueCI 60.5% - 65.4%

SkillsBench

VALS-AI · Professional reasoning · Objective

Applied professional skills tasks.

Rank #20 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 51.3%
Percentile: 42.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: skillsbench; provider: DeepSeek.

42.4% percentile inside its fair comparison set

51.3%Raw benchmark valueCI 42.3% - 60.3%

TaxEval v2

VALS-AI · Professional reasoning · Objective

Answer quality on tax questions and responses.

Rank #66 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 72.1%
Percentile: 49.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: tax_eval_v2; provider: DeepSeek.

49.6% percentile inside its fair comparison set

72.1%Raw benchmark valueCI 70.4% - 73.8%

Data analysis

LB · Professional reasoning · Objective

Structured data manipulation and table reasoning accuracy.

Rank #37

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 74.5%
Percentile: 44.6%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category: Data Analysis. Tasks scored: 3.

44.6% percentile inside its fair comparison set

74.5%Raw benchmark value

Overall

LB · Professional reasoning · Objective

Average objective performance across LiveBench's current public category mix.

Rank #53

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 71.6%
Percentile: 20%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Category averages included: 7.

20% percentile inside its fair comparison set

71.6%Raw benchmark value

Consecutive events

LB · Professional reasoning · Objective

Objective consecutive events score in LiveBench.

Rank #42

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 75.4%
Percentile: 36.9%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: consecutive_events. Category: Data Analysis.

36.9% percentile inside its fair comparison set

75.4%Raw benchmark value

Table join

LB · Professional reasoning · Objective

Objective table join score in LiveBench.

Rank #33

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 48.2%
Percentile: 50.8%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: tablejoin. Category: Data Analysis.

50.8% percentile inside its fair comparison set

48.2%Raw benchmark value

Table reformat

LB · Professional reasoning · Objective

Objective table reformat score in LiveBench.

Rank #1

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 100%
Percentile: 100%
Last updated: stale
Eligibility: headline eligible

Derived from the official LiveBench website leaderboard table. Task: tablereformat. Category: Data Analysis.

100% percentile inside its fair comparison set

100%Raw benchmark value

Public Benefits Bench v1

VALS-AI · Professional reasoning · Objective

Answering public-benefits questions across the benefits lifecycle.

Rank #7 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 57.6%
Percentile: 50%
Last updated: stale
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: public-benefits-bench-v1; provider: DeepSeek.

50% percentile inside its fair comparison set

57.6%Raw benchmark valueCI 55.1% - 60.1%

Hallucination

BB · Professional reasoning · Rubric

Accuracy and fabrication resistance under prompts that invite unsupported claims.

Rank #35 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 51.1%
Percentile: 0%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-hallucination.

0% percentile inside its fair comparison set

51.1%Raw benchmark value

BS pushback

BB · Professional reasoning · Rubric

Resistance to confidently accepting bogus assumptions in expert-style prompts.

Rank #19 · Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: BridgeBench
Raw value: 36%
Percentile: 5.6%
Last updated: archived
Eligibility: benchmark_derived_model

Parsed from the BridgeBench page for bridgebench-pushback.

5.6% percentile inside its fair comparison set

36%Raw benchmark value

Long context2 benchmarks71.8%

Long Context Reasoning

AA · Long context · Objective

It checks whether long-context claims survive contact with retrieval, memory, or long-document tasks.

Rank #49 · Source label: DeepSeek V4 Pro 0813 (Max)

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 80.3%
Percentile: 89.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Artificial Analysis public leaderboard field `lcr`.

89.7% percentile inside its fair comparison set

80.3%Raw benchmark value

CorpFin v2

VALS-AI · Long context · Objective

It checks whether long-context claims survive contact with retrieval, memory, or long-document tasks.

Rank #55 · Source label: deepseek/deepseek-v4-pro

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Vals AI
Raw value: 61.4%
Percentile: 53.8%
Last updated: aging
Eligibility: headline eligible

Parsed from Vals AI BenchmarkView overall scores. Vals slug: corp_fin_v2; provider: DeepSeek.

53.8% percentile inside its fair comparison set

61.4%Raw benchmark valueCI 59.5% - 63.3%

Multilingual18 benchmarksn/a

Text Arena · Chinese

AR · Multilingual · Human

Observed user preference in Arena's Text Arena chinese leaderboard.

Rank #73

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,489
Percentile: 79.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: chinese. Source rank: #79. Votes: 3384. Organization: deepseek. License: MIT.

79.5% percentile inside its fair comparison set

1,489Raw benchmark valueCI 1,478 - 1,499

Text Arena · French

AR · Multilingual · Human

Observed user preference in Arena's Text Arena french leaderboard.

Rank #42

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,485
Percentile: 84%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: french. Source rank: #46. Votes: 2205. Organization: deepseek. License: MIT.

84% percentile inside its fair comparison set

1,485Raw benchmark valueCI 1,470 - 1,499

Text Arena · German

AR · Multilingual · Human

Observed user preference in Arena's Text Arena german leaderboard.

Rank #39

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,468
Percentile: 86%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: german. Source rank: #43. Votes: 999. Organization: deepseek. License: MIT.

86% percentile inside its fair comparison set

1,468Raw benchmark valueCI 1,449 - 1,488

Text Arena · Japanese

AR · Multilingual · Human

Observed user preference in Arena's Text Arena japanese leaderboard.

Rank #29

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,457
Percentile: 88.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: japanese. Source rank: #31. Votes: 715. Organization: deepseek. License: MIT.

88.5% percentile inside its fair comparison set

1,457Raw benchmark valueCI 1,433 - 1,481

Text Arena · Korean

AR · Multilingual · Human

Observed user preference in Arena's Text Arena korean leaderboard.

Rank #49

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,419
Percentile: 80.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: korean. Source rank: #56. Votes: 1106. Organization: deepseek. License: MIT.

80.3% percentile inside its fair comparison set

1,419Raw benchmark valueCI 1,400 - 1,438

Text Arena · Non English

AR · Multilingual · Human

Observed user preference in Arena's Text Arena non_english leaderboard.

Rank #57

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,445
Percentile: 85.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: non_english. Source rank: #63. Votes: 32269. Organization: deepseek. License: MIT.

85.1% percentile inside its fair comparison set

1,445Raw benchmark valueCI 1,440 - 1,450

Text Arena · Polish

AR · Multilingual · Human

Observed user preference in Arena's Text Arena polish leaderboard.

Rank #38

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,466
Percentile: 81.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: polish. Source rank: #45. Votes: 1123. Organization: deepseek. License: MIT.

81.4% percentile inside its fair comparison set

1,466Raw benchmark valueCI 1,448 - 1,484

Text Arena · Russian

AR · Multilingual · Human

Observed user preference in Arena's Text Arena russian leaderboard.

Rank #53

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,460
Percentile: 84.8%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: russian. Source rank: #59. Votes: 6313. Organization: deepseek. License: MIT.

84.8% percentile inside its fair comparison set

1,460Raw benchmark valueCI 1,452 - 1,469

Text Arena · Spanish

AR · Multilingual · Human

Observed user preference in Arena's Text Arena spanish leaderboard.

Rank #51

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,458
Percentile: 80.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: spanish. Source rank: #56. Votes: 1726. Organization: deepseek. License: MIT.

80.7% percentile inside its fair comparison set

1,458Raw benchmark valueCI 1,442 - 1,473

Text Arena · Chinese · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena chinese leaderboard.

Rank #69

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,481
Percentile: 80.7%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: chinese. Source rank: #72. Votes: 3384. Organization: deepseek. License: MIT.

80.7% percentile inside its fair comparison set

1,481Raw benchmark valueCI 1,470 - 1,492

Text Arena · French · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena french leaderboard.

Rank #38

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,472
Percentile: 85.6%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: french. Source rank: #40. Votes: 2205. Organization: deepseek. License: MIT.

85.6% percentile inside its fair comparison set

1,472Raw benchmark valueCI 1,457 - 1,486

Text Arena · German · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena german leaderboard.

Rank #39

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,457
Percentile: 86%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: german. Source rank: #43. Votes: 999. Organization: deepseek. License: MIT.

86% percentile inside its fair comparison set

1,457Raw benchmark valueCI 1,437 - 1,476

Text Arena · Japanese · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena japanese leaderboard.

Rank #29

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,447
Percentile: 88.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: japanese. Source rank: #32. Votes: 715. Organization: deepseek. License: MIT.

88.5% percentile inside its fair comparison set

1,447Raw benchmark valueCI 1,423 - 1,470

Text Arena · Korean · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena korean leaderboard.

Rank #47

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,411
Percentile: 81.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: korean. Source rank: #53. Votes: 1106. Organization: deepseek. License: MIT.

81.1% percentile inside its fair comparison set

1,411Raw benchmark valueCI 1,392 - 1,430

Text Arena · Non English · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena non_english leaderboard.

Rank #48

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,439
Percentile: 87.5%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: non_english. Source rank: #52. Votes: 32269. Organization: deepseek. License: MIT.

87.5% percentile inside its fair comparison set

1,439Raw benchmark valueCI 1,434 - 1,443

Text Arena · Polish · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena polish leaderboard.

Rank #30

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,462
Percentile: 85.4%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: polish. Source rank: #32. Votes: 1123. Organization: deepseek. License: MIT.

85.4% percentile inside its fair comparison set

1,462Raw benchmark valueCI 1,444 - 1,479

Text Arena · Russian · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena russian leaderboard.

Rank #41

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,453
Percentile: 88.3%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: russian. Source rank: #46. Votes: 6313. Organization: deepseek. License: MIT.

88.3% percentile inside its fair comparison set

1,453Raw benchmark valueCI 1,445 - 1,461

Text Arena · Spanish · No Style Control

AR · Multilingual · Human

Observed user preference in Arena's Text Arena spanish leaderboard.

Rank #37

verified runtimeexact alias

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,458
Percentile: 86.1%
Last updated: recent
Eligibility: headline eligible

Parsed from Arena leaderboard dataset row `deepseek-v4-pro`. Category: spanish. Source rank: #38. Votes: 1726. Organization: deepseek. License: MIT.

86.1% percentile inside its fair comparison set

1,458Raw benchmark valueCI 1,442 - 1,473

Source links and registry checks

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Arena

Oct 3, 2026

source →

official

Artificial Analysis

Oct 3, 2026

source →

official

LiveBench

Oct 3, 2026

source →

deepseek-v4-pro

Current snapshot.

Source-linked scores by benchmark

Source links and registry checks

Loading model evidence.

deepseek-v4-pro

Current snapshot.

Source-linked scores by benchmark

Source links and registry checks