Model profile · DeepSeek

deepseek-v4-flash

Unknown weightsmid · registry tag 2026 benchmark-derived

Thin verified coverage

Reads as thin verified coverage across the resolved source data.

Visible coverage: 36.3%
Verified coverage: 36.3%
Spread: n/a
Last verified: Jun 20, 2026

textcodedocument4 aliases29 official source links

Open compare

Data version

Current snapshot.

Data version Jun 20, 2026Model list checked9 providers · 1081 tracked modelsPage refreshed Jul 5, 2026

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

Source-linked scores by benchmark

Each row keeps the benchmark source, source type, raw metric, and percentile inside its fair comparison set.

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

Chat / text38 benchmarks74.8%

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 #23 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 37
Percentile: 94.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `intelligenceIndex`.

94.4% percentile inside its fair comparison set

37Raw 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 #25 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 35.5%
Percentile: 91.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `omniscienceAccuracy`.

91.9% percentile inside its fair comparison set

35.5%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 #282 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 4.2%
Percentile: 5.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `omniscienceNonHallucination`.

5.7% percentile inside its fair comparison set

4.2%Raw benchmark value

IFBench

AA · Chat / text · Objective

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

Rank #16 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 73.5%
Percentile: 95.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `ifbench`.

95.2% percentile inside its fair comparison set

73.5%Raw benchmark value

Blended 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 #65 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: $0.2 /1M tokens
Percentile: 77.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `price1mBlended0To3To1`.

77.2% percentile inside its fair comparison set

$0.2 /1M tokensRaw 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 #76 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: $0.1 /1M input tokens
Percentile: 73.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `price1mInputTokens`.

73.9% percentile inside its fair comparison set

$0.1 /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 #55 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: $0.3 /1M output tokens
Percentile: 80.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `price1mOutputTokens`.

80.4% percentile inside its fair comparison set

$0.3 /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 #86 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 110.2 tokens/s
Percentile: 59.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `medianOutputTokensPerSecond`.

59.5% percentile inside its fair comparison set

110.2 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 #62 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 1.45s
Percentile: 71%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `medianTimeToFirstTokenSeconds`.

71% percentile inside its fair comparison set

1.45sRaw 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 #189 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 52.36s
Percentile: 10.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `medianTimeToFirstAnswerTokenSeconds`.

10.5% percentile inside its fair comparison set

52.36sRaw benchmark value

Openness Index

AA · Chat / text · Combined

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

Rank #34 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 50
Percentile: 83.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `opennessBreakdown.opennessIndex`.

83.9% percentile inside its fair comparison set

50Raw 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 #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,434
Percentile: 84.3%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: overall. Source rank: #67. Votes: 28291. Organization: deepseek. License: MIT.

84.3% percentile inside its fair comparison set

1,434Raw benchmark valueCI 1,429 - 1,439

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 #46

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,409
Percentile: 86.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: creative_writing. Source rank: #60. Votes: 4490. Organization: deepseek. License: MIT.

86.1% percentile inside its fair comparison set

1,409Raw benchmark valueCI 1,400 - 1,419

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 #57

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,442
Percentile: 82.8%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: english. Source rank: #71. Votes: 13110. Organization: deepseek. License: MIT.

82.8% percentile inside its fair comparison set

1,442Raw benchmark valueCI 1,435 - 1,448

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 #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,428
Percentile: 84.3%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: exclude_ties. Source rank: #67. Votes: 21507. Organization: deepseek. License: MIT.

84.3% percentile inside its fair comparison set

1,428Raw benchmark valueCI 1,422 - 1,434

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 #51

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,457
Percentile: 84.6%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: hard_prompts. Source rank: #65. Votes: 18735. Organization: deepseek. License: MIT.

84.6% percentile inside its fair comparison set

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

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 #53

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,460
Percentile: 84%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: hard_prompts_english. Source rank: #66. Votes: 9228. Organization: deepseek. License: MIT.

84% percentile inside its fair comparison set

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

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 #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,428
Percentile: 84.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: instruction_following. Source rank: #62. Votes: 9473. Organization: deepseek. License: MIT.

84.9% percentile inside its fair comparison set

1,428Raw benchmark valueCI 1,421 - 1,435

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 #49

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,446
Percentile: 84.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: longer_query. Source rank: #62. Votes: 12189. Organization: deepseek. License: MIT.

84.2% percentile inside its fair comparison set

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

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 aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,452
Percentile: 86.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: multi_turn. Source rank: #56. Votes: 5118. Organization: deepseek. License: MIT.

86.7% percentile inside its fair comparison set

1,452Raw benchmark valueCI 1,442 - 1,461

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 #47

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,430
Percentile: 85.8%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: overall. Source rank: #59. Votes: 28291. Organization: deepseek. License: MIT.

85.8% percentile inside its fair comparison set

1,430Raw benchmark valueCI 1,425 - 1,434

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 #45

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,405
Percentile: 86.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: creative_writing. Source rank: #58. Votes: 4490. Organization: deepseek. License: MIT.

86.4% percentile inside its fair comparison set

1,405Raw benchmark valueCI 1,395 - 1,414

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 #54

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,437
Percentile: 83.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: english. Source rank: #66. Votes: 13110. Organization: deepseek. License: MIT.

83.7% percentile inside its fair comparison set

1,437Raw benchmark valueCI 1,431 - 1,443

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 #49

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,421
Percentile: 85.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: exclude_ties. Source rank: #61. Votes: 21507. Organization: deepseek. License: MIT.

85.2% percentile inside its fair comparison set

1,421Raw benchmark valueCI 1,415 - 1,427

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 #45

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,441
Percentile: 86.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: hard_prompts. Source rank: #55. Votes: 18735. Organization: deepseek. License: MIT.

86.5% percentile inside its fair comparison set

1,441Raw benchmark valueCI 1,436 - 1,447

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 #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,444
Percentile: 84.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: hard_prompts_english. Source rank: #59. Votes: 9228. Organization: deepseek. License: MIT.

84.9% percentile inside its fair comparison set

1,444Raw benchmark valueCI 1,437 - 1,451

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 #42

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,419
Percentile: 87.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: instruction_following. Source rank: #50. Votes: 9473. Organization: deepseek. License: MIT.

87.4% percentile inside its fair comparison set

1,419Raw benchmark valueCI 1,412 - 1,426

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 #43

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,432
Percentile: 86.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: longer_query. Source rank: #51. Votes: 12189. Organization: deepseek. License: MIT.

86.2% percentile inside its fair comparison set

1,432Raw benchmark valueCI 1,426 - 1,439

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 aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,447
Percentile: 89.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: multi_turn. Source rank: #45. Votes: 5118. Organization: deepseek. License: MIT.

89.5% percentile inside its fair comparison set

1,447Raw benchmark valueCI 1,438 - 1,456

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 #25

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 63.1%
Percentile: 77.8%
Last updated: archived
Eligibility: benchmark_derived_model

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

77.8% percentile inside its fair comparison set

63.1%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 #65

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 70.1%
Percentile: 40.7%
Last updated: archived
Eligibility: benchmark_derived_model

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

40.7% percentile inside its fair comparison set

70.1%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 #21

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 62.4%
Percentile: 81.5%
Last updated: archived
Eligibility: benchmark_derived_model

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

81.5% percentile inside its fair comparison set

62.4%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 #43

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 54.1%
Percentile: 61.1%
Last updated: archived
Eligibility: benchmark_derived_model

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

61.1% percentile inside its fair comparison set

54.1%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 #24

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 67.3%
Percentile: 78.7%
Last updated: archived
Eligibility: benchmark_derived_model

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

78.7% percentile inside its fair comparison set

67.3%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 #14

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 68.9%
Percentile: 88%
Last updated: archived
Eligibility: benchmark_derived_model

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

88% percentile inside its fair comparison set

68.9%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 #55

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 89.5%
Percentile: 50%
Last updated: archived
Eligibility: benchmark_derived_model

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

50% percentile inside its fair comparison set

89.5%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 #69

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 46.9%
Percentile: 37%
Last updated: archived
Eligibility: benchmark_derived_model

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

37% percentile inside its fair comparison set

46.9%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 #59

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 74%
Percentile: 45.8%
Last updated: archived
Eligibility: benchmark_derived_model

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

45.8% percentile inside its fair comparison set

74%Raw benchmark value

Coding13 benchmarks66.3%

Terminal-Bench Hard

AA · Coding · Objective

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

Rank #35 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 35.6%
Percentile: 89.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `terminalbenchHard`.

89.4% percentile inside its fair comparison set

35.6%Raw benchmark value

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 #41 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 42%
Percentile: 89.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `scicode`.

89.1% percentile inside its fair comparison set

42%Raw benchmark value

Coding Index

AA · Coding · Combined

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

Rank #16 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 56
Percentile: 80%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `codingIndex`.

80% percentile inside its fair comparison set

56Raw benchmark value

Agentic Index

AA · Coding · Combined

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

Rank #10 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 31
Percentile: 80.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `agenticIndex`.

80.4% percentile inside its fair comparison set

31Raw 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 #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,480
Percentile: 84.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: coding. Source rank: #66. Votes: 8206. Organization: deepseek. License: MIT.

84.1% percentile inside its fair comparison set

1,480Raw benchmark valueCI 1,472 - 1,487

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 #49

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,452
Percentile: 85%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: coding. Source rank: #59. Votes: 8206. Organization: deepseek. License: MIT.

85% percentile inside its fair comparison set

1,452Raw benchmark valueCI 1,444 - 1,459

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 #46

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 50%
Percentile: 60.2%
Last updated: archived
Eligibility: benchmark_derived_model

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

60.2% percentile inside its fair comparison set

50%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 #78

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 69.2%
Percentile: 28.7%
Last updated: archived
Eligibility: benchmark_derived_model

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

28.7% percentile inside its fair comparison set

69.2%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 #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 40%
Percentile: 58.3%
Last updated: archived
Eligibility: benchmark_derived_model

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

58.3% percentile inside its fair comparison set

40%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 #42

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 40%
Percentile: 64.5%
Last updated: archived
Eligibility: benchmark_derived_model

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

64.5% percentile inside its fair comparison set

40%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 #34

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 70%
Percentile: 77.8%
Last updated: archived
Eligibility: benchmark_derived_model

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

77.8% percentile inside its fair comparison set

70%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 #69

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 73.2%
Percentile: 38%
Last updated: archived
Eligibility: benchmark_derived_model

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

38% percentile inside its fair comparison set

73.2%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 #82

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 65.2%
Percentile: 25.9%
Last updated: archived
Eligibility: benchmark_derived_model

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

25.9% percentile inside its fair comparison set

65.2%Raw benchmark value

Reasoning / math / science15 benchmarks73.2%

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 #17 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 27.8%
Percentile: 95.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `hle`.

95.7% percentile inside its fair comparison set

27.8%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 #21 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 86.7%
Percentile: 94.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `gpqa`.

94.7% percentile inside its fair comparison set

86.7%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 #24 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 3.4%
Percentile: 92.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `critpt`.

92.4% percentile inside its fair comparison set

3.4%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 #61

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,427
Percentile: 80.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: math. Source rank: #77. Votes: 1523. Organization: deepseek. License: MIT.

80.9% percentile inside its fair comparison set

1,427Raw benchmark valueCI 1,411 - 1,442

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 #56

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,428
Percentile: 82.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: math. Source rank: #68. Votes: 1523. Organization: deepseek. License: MIT.

82.5% percentile inside its fair comparison set

1,428Raw benchmark valueCI 1,412 - 1,443

Mathematics

LB · Reasoning / math / science · Objective

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

Rank #48

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 79.6%
Percentile: 56.5%
Last updated: archived
Eligibility: benchmark_derived_model

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

56.5% percentile inside its fair comparison set

79.6%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 #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 70.6%
Percentile: 52.8%
Last updated: archived
Eligibility: benchmark_derived_model

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

52.8% percentile inside its fair comparison set

70.6%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 #14

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 98%
Percentile: 94.4%
Last updated: archived
Eligibility: benchmark_derived_model

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

94.4% 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 #60

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 37%
Percentile: 45.4%
Last updated: archived
Eligibility: benchmark_derived_model

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

45.4% percentile inside its fair comparison set

37%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 #3

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 97.1%
Percentile: 99.1%
Last updated: archived
Eligibility: benchmark_derived_model

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

99.1% percentile inside its fair comparison set

97.1%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 #46

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 86.5%
Percentile: 58.3%
Last updated: archived
Eligibility: benchmark_derived_model

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

58.3% percentile inside its fair comparison set

86.5%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 #36

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 73.1%
Percentile: 69.4%
Last updated: archived
Eligibility: benchmark_derived_model

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

69.4% percentile inside its fair comparison set

73.1%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 #55

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 49.3%
Percentile: 49.5%
Last updated: archived
Eligibility: benchmark_derived_model

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

49.5% percentile inside its fair comparison set

49.3%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 #35

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 96%
Percentile: 73.1%
Last updated: archived
Eligibility: benchmark_derived_model

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

73.1% percentile inside its fair comparison set

96%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 #53

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 64%
Percentile: 53.7%
Last updated: archived
Eligibility: benchmark_derived_model

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

53.7% percentile inside its fair comparison set

64%Raw benchmark value

Professional reasoning24 benchmarks79.9%

GDPval-AA

AA · Professional reasoning · Rubric

Agentic performance on economically valuable work tasks.

Rank #16 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 1,189
Percentile: 67.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `gdpvalBreakdown.elo`.

67.4% percentile inside its fair comparison set

1,189Raw benchmark value

Text Arena · Expert

AR · Professional reasoning · Human

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

Rank #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,458
Percentile: 81.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: expert. Source rank: #67. Votes: 2633. Organization: deepseek. License: MIT.

81.5% percentile inside its fair comparison set

1,458Raw benchmark valueCI 1,446 - 1,470

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 #57

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,429
Percentile: 82.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_business_and_management_and_financial_operations. Source rank: #73. Votes: 5618. Organization: deepseek. License: MIT.

82.4% percentile inside its fair comparison set

1,429Raw benchmark valueCI 1,420 - 1,437

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 #53

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,405
Percentile: 83.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_entertainment_and_sports_and_media. Source rank: #66. Votes: 5818. Organization: deepseek. License: MIT.

83.9% percentile inside its fair comparison set

1,405Raw benchmark valueCI 1,396 - 1,414

Text Arena · Industry Legal And Government

AR · Professional reasoning · Human

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

Rank #55

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,441
Percentile: 81.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_legal_and_government. Source rank: #73. Votes: 2174. Organization: deepseek. License: MIT.

81.9% percentile inside its fair comparison set

1,441Raw benchmark valueCI 1,427 - 1,454

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 #45

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,459
Percentile: 86.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_life_and_physical_and_social_science. Source rank: #57. Votes: 4607. Organization: deepseek. License: MIT.

86.4% percentile inside its fair comparison set

1,459Raw benchmark valueCI 1,449 - 1,468

Text Arena · Industry Mathematical

AR · Professional reasoning · Human

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

Rank #62

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,437
Percentile: 80.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_mathematical. Source rank: #75. Votes: 1591. Organization: deepseek. License: MIT.

80.2% percentile inside its fair comparison set

1,437Raw benchmark valueCI 1,422 - 1,453

Text Arena · Industry Medicine And Healthcare

AR · Professional reasoning · Human

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

Rank #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,458
Percentile: 82.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_medicine_and_healthcare. Source rank: #66. Votes: 2012. Organization: deepseek. License: MIT.

82.7% percentile inside its fair comparison set

1,458Raw benchmark valueCI 1,444 - 1,472

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 #56

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,468
Percentile: 83.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_software_and_it_services. Source rank: #70. Votes: 11498. Organization: deepseek. License: MIT.

83.1% percentile inside its fair comparison set

1,468Raw benchmark valueCI 1,462 - 1,475

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 #48

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,419
Percentile: 85.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_writing_and_literature_and_language. Source rank: #61. Votes: 6667. Organization: deepseek. License: MIT.

85.5% percentile inside its fair comparison set

1,419Raw benchmark valueCI 1,411 - 1,427

Text Arena · Expert · No Style Control

AR · Professional reasoning · Human

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

Rank #48

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,441
Percentile: 82.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: expert. Source rank: #57. Votes: 2633. Organization: deepseek. License: MIT.

82.9% percentile inside its fair comparison set

1,441Raw benchmark valueCI 1,429 - 1,453

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 #60

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,416
Percentile: 81.4%
Last updated: recent
Eligibility: benchmark_derived_model

81.4% percentile inside its fair comparison set

1,416Raw benchmark valueCI 1,407 - 1,424

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 #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,400
Percentile: 84.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_entertainment_and_sports_and_media. Source rank: #63. Votes: 5818. Organization: deepseek. License: MIT.

84.2% percentile inside its fair comparison set

1,400Raw benchmark valueCI 1,391 - 1,408

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 #59

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,431
Percentile: 80.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_legal_and_government. Source rank: #72. Votes: 2174. Organization: deepseek. License: MIT.

80.5% percentile inside its fair comparison set

1,431Raw benchmark valueCI 1,418 - 1,445

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 #43

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,449
Percentile: 87%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_life_and_physical_and_social_science. Source rank: #51. Votes: 4607. Organization: deepseek. License: MIT.

87% percentile inside its fair comparison set

1,449Raw benchmark valueCI 1,439 - 1,458

Text Arena · Industry Mathematical · No Style Control

AR · Professional reasoning · Human

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

Rank #57

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,433
Percentile: 81.8%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_mathematical. Source rank: #67. Votes: 1591. Organization: deepseek. License: MIT.

81.8% percentile inside its fair comparison set

1,433Raw benchmark valueCI 1,418 - 1,449

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 #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,448
Percentile: 83.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_medicine_and_healthcare. Source rank: #58. Votes: 2012. Organization: deepseek. License: MIT.

83.4% percentile inside its fair comparison set

1,448Raw benchmark valueCI 1,434 - 1,462

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 #51

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,449
Percentile: 84.6%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_software_and_it_services. Source rank: #63. Votes: 11498. Organization: deepseek. License: MIT.

84.6% percentile inside its fair comparison set

1,449Raw benchmark valueCI 1,443 - 1,456

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 #41

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,413
Percentile: 87.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_writing_and_literature_and_language. Source rank: #52. Votes: 6667. Organization: deepseek. License: MIT.

87.7% percentile inside its fair comparison set

1,413Raw benchmark valueCI 1,405 - 1,421

Data analysis

LB · Professional reasoning · Objective

Structured data manipulation and table reasoning accuracy.

Rank #34

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 68%
Percentile: 69.4%
Last updated: archived
Eligibility: benchmark_derived_model

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

69.4% percentile inside its fair comparison set

68%Raw benchmark value

Overall

LB · Professional reasoning · Objective

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

Rank #45

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 67.2%
Percentile: 59.3%
Last updated: archived
Eligibility: benchmark_derived_model

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

59.3% percentile inside its fair comparison set

67.2%Raw benchmark value

Consecutive events

LB · Professional reasoning · Objective

Objective consecutive events score in LiveBench.

Rank #35

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 59.9%
Percentile: 68.5%
Last updated: archived
Eligibility: benchmark_derived_model

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

68.5% percentile inside its fair comparison set

59.9%Raw benchmark value

Table join

LB · Professional reasoning · Objective

Objective table join score in LiveBench.

Rank #24

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 46.1%
Percentile: 78.7%
Last updated: archived
Eligibility: benchmark_derived_model

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

78.7% percentile inside its fair comparison set

46.1%Raw benchmark value

Table reformat

LB · Professional reasoning · Objective

Objective table reformat score in LiveBench.

Rank #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 98%
Percentile: 74.1%
Last updated: archived
Eligibility: benchmark_derived_model

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

74.1% percentile inside its fair comparison set

98%Raw benchmark value

Search / tool use1 benchmark95.1%

Tau2-Bench Telecom

AA · Search / tool use · Objective

It matters when the model must browse, call tools, and recover useful answers from external systems.

Rank #16 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 95%
Percentile: 95.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `tau2`.

95.1% percentile inside its fair comparison set

95%Raw benchmark value

Long context1 benchmark87.6%

Long Context Reasoning

AA · Long context · Objective

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

Rank #40 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 62.7%
Percentile: 87.6%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `lcr`.

87.6% percentile inside its fair comparison set

62.7%Raw benchmark value

Multilingual14 benchmarks80%

Text Arena · Chinese

AR · Multilingual · Human

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

Rank #51

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,478
Percentile: 83.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: chinese. Source rank: #63. Votes: 1386. Organization: deepseek. License: MIT.

83.1% percentile inside its fair comparison set

1,478Raw benchmark valueCI 1,461 - 1,495

Text Arena · French

AR · Multilingual · Human

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

Rank #55

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,450
Percentile: 75%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: french. Source rank: #70. Votes: 1077. Organization: deepseek. License: MIT.

75% percentile inside its fair comparison set

1,450Raw benchmark valueCI 1,429 - 1,470

Text Arena · German

AR · Multilingual · Human

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

Rank #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,425
Percentile: 78.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: german. Source rank: #69. Votes: 440. Organization: deepseek. License: MIT.

78.5% percentile inside its fair comparison set

1,425Raw benchmark valueCI 1,397 - 1,454

Text Arena · Japanese

AR · Multilingual · Human

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

Rank #40

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,398
Percentile: 80.8%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: japanese. Source rank: #56. Votes: 285. Organization: deepseek. License: MIT.

80.8% percentile inside its fair comparison set

1,398Raw benchmark valueCI 1,360 - 1,436

Text Arena · Korean

AR · Multilingual · Human

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

Rank #44

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,392
Percentile: 79.3%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: korean. Source rank: #57. Votes: 431. Organization: deepseek. License: MIT.

79.3% percentile inside its fair comparison set

1,392Raw benchmark valueCI 1,362 - 1,422

Text Arena · Russian

AR · Multilingual · Human

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

Rank #49

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,437
Percentile: 83.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: russian. Source rank: #63. Votes: 3180. Organization: deepseek. License: MIT.

83.4% percentile inside its fair comparison set

1,437Raw benchmark valueCI 1,426 - 1,448

Text Arena · Spanish

AR · Multilingual · Human

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

Rank #45

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,437
Percentile: 79.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: spanish. Source rank: #57. Votes: 831. Organization: deepseek. License: MIT.

79.4% percentile inside its fair comparison set

1,437Raw benchmark valueCI 1,415 - 1,458

Text Arena · Chinese · No Style Control

AR · Multilingual · Human

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

Rank #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,468
Percentile: 83.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: chinese. Source rank: #60. Votes: 1386. Organization: deepseek. License: MIT.

83.4% percentile inside its fair comparison set

1,468Raw benchmark valueCI 1,451 - 1,485

Text Arena · French · No Style Control

AR · Multilingual · Human

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

Rank #57

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,438
Percentile: 74.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: french. Source rank: #69. Votes: 1077. Organization: deepseek. License: MIT.

74.1% percentile inside its fair comparison set

1,438Raw benchmark valueCI 1,418 - 1,459

Text Arena · German · No Style Control

AR · Multilingual · Human

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

Rank #53

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,420
Percentile: 78.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: german. Source rank: #66. Votes: 440. Organization: deepseek. License: MIT.

78.1% percentile inside its fair comparison set

1,420Raw benchmark valueCI 1,392 - 1,449

Text Arena · Japanese · No Style Control

AR · Multilingual · Human

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

Rank #43

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,385
Percentile: 79.3%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: japanese. Source rank: #56. Votes: 285. Organization: deepseek. License: MIT.

79.3% percentile inside its fair comparison set

1,385Raw benchmark valueCI 1,347 - 1,422

Text Arena · Korean · No Style Control

AR · Multilingual · Human

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

Rank #53

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,379
Percentile: 75%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: korean. Source rank: #63. Votes: 431. Organization: deepseek. License: MIT.

75% percentile inside its fair comparison set

1,379Raw benchmark valueCI 1,350 - 1,409

Text Arena · Russian · No Style Control

AR · Multilingual · Human

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

Rank #40

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,431
Percentile: 86.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: russian. Source rank: #49. Votes: 3180. Organization: deepseek. License: MIT.

86.5% percentile inside its fair comparison set

1,431Raw benchmark valueCI 1,420 - 1,442

Text Arena · Spanish · No Style Control

AR · Multilingual · Human

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

Rank #35

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,443
Percentile: 84.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: spanish. Source rank: #42. Votes: 831. Organization: deepseek. License: MIT.

84.1% percentile inside its fair comparison set

1,443Raw benchmark valueCI 1,422 - 1,465

Source links and registry checks

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Arena

Jun 20, 2026

source →

official

Artificial Analysis

Jun 20, 2026

source →

official

LiveBench

Jun 20, 2026

source →

Model profile · DeepSeek

deepseek-v4-flash

Unknown weightsmid · registry tag 2026 benchmark-derived

Thin verified coverage

Reads as thin verified coverage across the resolved source data.

Visible coverage: 36.3%
Verified coverage: 36.3%
Spread: n/a
Last verified: Jun 20, 2026

textcodedocument4 aliases29 official source links

Open compare

Data version

Current snapshot.

Data version Jun 20, 2026Model list checked9 providers · 1081 tracked modelsPage refreshed Jul 5, 2026

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

Source-linked scores by benchmark

Each row keeps the benchmark source, source type, raw metric, and percentile inside its fair comparison set.

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

Chat / text38 benchmarks74.8%

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 #23 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 37
Percentile: 94.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `intelligenceIndex`.

94.4% percentile inside its fair comparison set

37Raw 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 #25 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 35.5%
Percentile: 91.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `omniscienceAccuracy`.

91.9% percentile inside its fair comparison set

35.5%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 #282 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 4.2%
Percentile: 5.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `omniscienceNonHallucination`.

5.7% percentile inside its fair comparison set

4.2%Raw benchmark value

IFBench

AA · Chat / text · Objective

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

Rank #16 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 73.5%
Percentile: 95.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `ifbench`.

95.2% percentile inside its fair comparison set

73.5%Raw benchmark value

Blended 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 #65 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: $0.2 /1M tokens
Percentile: 77.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `price1mBlended0To3To1`.

77.2% percentile inside its fair comparison set

$0.2 /1M tokensRaw 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 #76 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: $0.1 /1M input tokens
Percentile: 73.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `price1mInputTokens`.

73.9% percentile inside its fair comparison set

$0.1 /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 #55 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: $0.3 /1M output tokens
Percentile: 80.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `price1mOutputTokens`.

80.4% percentile inside its fair comparison set

$0.3 /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 #86 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 110.2 tokens/s
Percentile: 59.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `medianOutputTokensPerSecond`.

59.5% percentile inside its fair comparison set

110.2 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 #62 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 1.45s
Percentile: 71%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `medianTimeToFirstTokenSeconds`.

71% percentile inside its fair comparison set

1.45sRaw 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 #189 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 52.36s
Percentile: 10.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `medianTimeToFirstAnswerTokenSeconds`.

10.5% percentile inside its fair comparison set

52.36sRaw benchmark value

Openness Index

AA · Chat / text · Combined

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

Rank #34 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 50
Percentile: 83.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `opennessBreakdown.opennessIndex`.

83.9% percentile inside its fair comparison set

50Raw 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 #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,434
Percentile: 84.3%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: overall. Source rank: #67. Votes: 28291. Organization: deepseek. License: MIT.

84.3% percentile inside its fair comparison set

1,434Raw benchmark valueCI 1,429 - 1,439

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 #46

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,409
Percentile: 86.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: creative_writing. Source rank: #60. Votes: 4490. Organization: deepseek. License: MIT.

86.1% percentile inside its fair comparison set

1,409Raw benchmark valueCI 1,400 - 1,419

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 #57

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,442
Percentile: 82.8%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: english. Source rank: #71. Votes: 13110. Organization: deepseek. License: MIT.

82.8% percentile inside its fair comparison set

1,442Raw benchmark valueCI 1,435 - 1,448

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 #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,428
Percentile: 84.3%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: exclude_ties. Source rank: #67. Votes: 21507. Organization: deepseek. License: MIT.

84.3% percentile inside its fair comparison set

1,428Raw benchmark valueCI 1,422 - 1,434

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 #51

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,457
Percentile: 84.6%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: hard_prompts. Source rank: #65. Votes: 18735. Organization: deepseek. License: MIT.

84.6% percentile inside its fair comparison set

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

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 #53

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,460
Percentile: 84%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: hard_prompts_english. Source rank: #66. Votes: 9228. Organization: deepseek. License: MIT.

84% percentile inside its fair comparison set

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

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 #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,428
Percentile: 84.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: instruction_following. Source rank: #62. Votes: 9473. Organization: deepseek. License: MIT.

84.9% percentile inside its fair comparison set

1,428Raw benchmark valueCI 1,421 - 1,435

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 #49

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,446
Percentile: 84.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: longer_query. Source rank: #62. Votes: 12189. Organization: deepseek. License: MIT.

84.2% percentile inside its fair comparison set

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

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 aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,452
Percentile: 86.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: multi_turn. Source rank: #56. Votes: 5118. Organization: deepseek. License: MIT.

86.7% percentile inside its fair comparison set

1,452Raw benchmark valueCI 1,442 - 1,461

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 #47

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,430
Percentile: 85.8%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: overall. Source rank: #59. Votes: 28291. Organization: deepseek. License: MIT.

85.8% percentile inside its fair comparison set

1,430Raw benchmark valueCI 1,425 - 1,434

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 #45

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,405
Percentile: 86.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: creative_writing. Source rank: #58. Votes: 4490. Organization: deepseek. License: MIT.

86.4% percentile inside its fair comparison set

1,405Raw benchmark valueCI 1,395 - 1,414

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 #54

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,437
Percentile: 83.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: english. Source rank: #66. Votes: 13110. Organization: deepseek. License: MIT.

83.7% percentile inside its fair comparison set

1,437Raw benchmark valueCI 1,431 - 1,443

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 #49

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,421
Percentile: 85.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: exclude_ties. Source rank: #61. Votes: 21507. Organization: deepseek. License: MIT.

85.2% percentile inside its fair comparison set

1,421Raw benchmark valueCI 1,415 - 1,427

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 #45

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,441
Percentile: 86.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: hard_prompts. Source rank: #55. Votes: 18735. Organization: deepseek. License: MIT.

86.5% percentile inside its fair comparison set

1,441Raw benchmark valueCI 1,436 - 1,447

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 #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,444
Percentile: 84.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: hard_prompts_english. Source rank: #59. Votes: 9228. Organization: deepseek. License: MIT.

84.9% percentile inside its fair comparison set

1,444Raw benchmark valueCI 1,437 - 1,451

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 #42

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,419
Percentile: 87.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: instruction_following. Source rank: #50. Votes: 9473. Organization: deepseek. License: MIT.

87.4% percentile inside its fair comparison set

1,419Raw benchmark valueCI 1,412 - 1,426

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 #43

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,432
Percentile: 86.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: longer_query. Source rank: #51. Votes: 12189. Organization: deepseek. License: MIT.

86.2% percentile inside its fair comparison set

1,432Raw benchmark valueCI 1,426 - 1,439

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 aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,447
Percentile: 89.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: multi_turn. Source rank: #45. Votes: 5118. Organization: deepseek. License: MIT.

89.5% percentile inside its fair comparison set

1,447Raw benchmark valueCI 1,438 - 1,456

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 #25

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 63.1%
Percentile: 77.8%
Last updated: archived
Eligibility: benchmark_derived_model

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

77.8% percentile inside its fair comparison set

63.1%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 #65

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 70.1%
Percentile: 40.7%
Last updated: archived
Eligibility: benchmark_derived_model

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

40.7% percentile inside its fair comparison set

70.1%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 #21

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 62.4%
Percentile: 81.5%
Last updated: archived
Eligibility: benchmark_derived_model

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

81.5% percentile inside its fair comparison set

62.4%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 #43

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 54.1%
Percentile: 61.1%
Last updated: archived
Eligibility: benchmark_derived_model

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

61.1% percentile inside its fair comparison set

54.1%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 #24

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 67.3%
Percentile: 78.7%
Last updated: archived
Eligibility: benchmark_derived_model

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

78.7% percentile inside its fair comparison set

67.3%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 #14

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 68.9%
Percentile: 88%
Last updated: archived
Eligibility: benchmark_derived_model

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

88% percentile inside its fair comparison set

68.9%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 #55

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 89.5%
Percentile: 50%
Last updated: archived
Eligibility: benchmark_derived_model

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

50% percentile inside its fair comparison set

89.5%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 #69

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 46.9%
Percentile: 37%
Last updated: archived
Eligibility: benchmark_derived_model

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

37% percentile inside its fair comparison set

46.9%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 #59

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 74%
Percentile: 45.8%
Last updated: archived
Eligibility: benchmark_derived_model

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

45.8% percentile inside its fair comparison set

74%Raw benchmark value

Coding13 benchmarks66.3%

Terminal-Bench Hard

AA · Coding · Objective

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

Rank #35 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 35.6%
Percentile: 89.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `terminalbenchHard`.

89.4% percentile inside its fair comparison set

35.6%Raw benchmark value

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 #41 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 42%
Percentile: 89.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `scicode`.

89.1% percentile inside its fair comparison set

42%Raw benchmark value

Coding Index

AA · Coding · Combined

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

Rank #16 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 56
Percentile: 80%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `codingIndex`.

80% percentile inside its fair comparison set

56Raw benchmark value

Agentic Index

AA · Coding · Combined

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

Rank #10 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 31
Percentile: 80.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `agenticIndex`.

80.4% percentile inside its fair comparison set

31Raw 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 #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,480
Percentile: 84.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: coding. Source rank: #66. Votes: 8206. Organization: deepseek. License: MIT.

84.1% percentile inside its fair comparison set

1,480Raw benchmark valueCI 1,472 - 1,487

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 #49

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,452
Percentile: 85%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: coding. Source rank: #59. Votes: 8206. Organization: deepseek. License: MIT.

85% percentile inside its fair comparison set

1,452Raw benchmark valueCI 1,444 - 1,459

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 #46

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 50%
Percentile: 60.2%
Last updated: archived
Eligibility: benchmark_derived_model

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

60.2% percentile inside its fair comparison set

50%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 #78

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 69.2%
Percentile: 28.7%
Last updated: archived
Eligibility: benchmark_derived_model

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

28.7% percentile inside its fair comparison set

69.2%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 #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 40%
Percentile: 58.3%
Last updated: archived
Eligibility: benchmark_derived_model

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

58.3% percentile inside its fair comparison set

40%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 #42

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 40%
Percentile: 64.5%
Last updated: archived
Eligibility: benchmark_derived_model

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

64.5% percentile inside its fair comparison set

40%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 #34

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 70%
Percentile: 77.8%
Last updated: archived
Eligibility: benchmark_derived_model

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

77.8% percentile inside its fair comparison set

70%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 #69

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 73.2%
Percentile: 38%
Last updated: archived
Eligibility: benchmark_derived_model

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

38% percentile inside its fair comparison set

73.2%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 #82

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 65.2%
Percentile: 25.9%
Last updated: archived
Eligibility: benchmark_derived_model

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

25.9% percentile inside its fair comparison set

65.2%Raw benchmark value

Reasoning / math / science15 benchmarks73.2%

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 #17 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 27.8%
Percentile: 95.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `hle`.

95.7% percentile inside its fair comparison set

27.8%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 #21 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 86.7%
Percentile: 94.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `gpqa`.

94.7% percentile inside its fair comparison set

86.7%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 #24 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 3.4%
Percentile: 92.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `critpt`.

92.4% percentile inside its fair comparison set

3.4%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 #61

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,427
Percentile: 80.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: math. Source rank: #77. Votes: 1523. Organization: deepseek. License: MIT.

80.9% percentile inside its fair comparison set

1,427Raw benchmark valueCI 1,411 - 1,442

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 #56

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,428
Percentile: 82.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: math. Source rank: #68. Votes: 1523. Organization: deepseek. License: MIT.

82.5% percentile inside its fair comparison set

1,428Raw benchmark valueCI 1,412 - 1,443

Mathematics

LB · Reasoning / math / science · Objective

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

Rank #48

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 79.6%
Percentile: 56.5%
Last updated: archived
Eligibility: benchmark_derived_model

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

56.5% percentile inside its fair comparison set

79.6%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 #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 70.6%
Percentile: 52.8%
Last updated: archived
Eligibility: benchmark_derived_model

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

52.8% percentile inside its fair comparison set

70.6%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 #14

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 98%
Percentile: 94.4%
Last updated: archived
Eligibility: benchmark_derived_model

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

94.4% 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 #60

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 37%
Percentile: 45.4%
Last updated: archived
Eligibility: benchmark_derived_model

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

45.4% percentile inside its fair comparison set

37%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 #3

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 97.1%
Percentile: 99.1%
Last updated: archived
Eligibility: benchmark_derived_model

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

99.1% percentile inside its fair comparison set

97.1%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 #46

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 86.5%
Percentile: 58.3%
Last updated: archived
Eligibility: benchmark_derived_model

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

58.3% percentile inside its fair comparison set

86.5%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 #36

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 73.1%
Percentile: 69.4%
Last updated: archived
Eligibility: benchmark_derived_model

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

69.4% percentile inside its fair comparison set

73.1%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 #55

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 49.3%
Percentile: 49.5%
Last updated: archived
Eligibility: benchmark_derived_model

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

49.5% percentile inside its fair comparison set

49.3%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 #35

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 96%
Percentile: 73.1%
Last updated: archived
Eligibility: benchmark_derived_model

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

73.1% percentile inside its fair comparison set

96%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 #53

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 64%
Percentile: 53.7%
Last updated: archived
Eligibility: benchmark_derived_model

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

53.7% percentile inside its fair comparison set

64%Raw benchmark value

Professional reasoning24 benchmarks79.9%

GDPval-AA

AA · Professional reasoning · Rubric

Agentic performance on economically valuable work tasks.

Rank #16 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 1,189
Percentile: 67.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `gdpvalBreakdown.elo`.

67.4% percentile inside its fair comparison set

1,189Raw benchmark value

Text Arena · Expert

AR · Professional reasoning · Human

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

Rank #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,458
Percentile: 81.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: expert. Source rank: #67. Votes: 2633. Organization: deepseek. License: MIT.

81.5% percentile inside its fair comparison set

1,458Raw benchmark valueCI 1,446 - 1,470

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 #57

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,429
Percentile: 82.4%
Last updated: recent
Eligibility: benchmark_derived_model

82.4% percentile inside its fair comparison set

1,429Raw benchmark valueCI 1,420 - 1,437

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 #53

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,405
Percentile: 83.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_entertainment_and_sports_and_media. Source rank: #66. Votes: 5818. Organization: deepseek. License: MIT.

83.9% percentile inside its fair comparison set

1,405Raw benchmark valueCI 1,396 - 1,414

Text Arena · Industry Legal And Government

AR · Professional reasoning · Human

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

Rank #55

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,441
Percentile: 81.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_legal_and_government. Source rank: #73. Votes: 2174. Organization: deepseek. License: MIT.

81.9% percentile inside its fair comparison set

1,441Raw benchmark valueCI 1,427 - 1,454

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 #45

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,459
Percentile: 86.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_life_and_physical_and_social_science. Source rank: #57. Votes: 4607. Organization: deepseek. License: MIT.

86.4% percentile inside its fair comparison set

1,459Raw benchmark valueCI 1,449 - 1,468

Text Arena · Industry Mathematical

AR · Professional reasoning · Human

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

Rank #62

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,437
Percentile: 80.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_mathematical. Source rank: #75. Votes: 1591. Organization: deepseek. License: MIT.

80.2% percentile inside its fair comparison set

1,437Raw benchmark valueCI 1,422 - 1,453

Text Arena · Industry Medicine And Healthcare

AR · Professional reasoning · Human

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

Rank #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,458
Percentile: 82.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_medicine_and_healthcare. Source rank: #66. Votes: 2012. Organization: deepseek. License: MIT.

82.7% percentile inside its fair comparison set

1,458Raw benchmark valueCI 1,444 - 1,472

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 #56

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,468
Percentile: 83.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_software_and_it_services. Source rank: #70. Votes: 11498. Organization: deepseek. License: MIT.

83.1% percentile inside its fair comparison set

1,468Raw benchmark valueCI 1,462 - 1,475

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 #48

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,419
Percentile: 85.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_writing_and_literature_and_language. Source rank: #61. Votes: 6667. Organization: deepseek. License: MIT.

85.5% percentile inside its fair comparison set

1,419Raw benchmark valueCI 1,411 - 1,427

Text Arena · Expert · No Style Control

AR · Professional reasoning · Human

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

Rank #48

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,441
Percentile: 82.9%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: expert. Source rank: #57. Votes: 2633. Organization: deepseek. License: MIT.

82.9% percentile inside its fair comparison set

1,441Raw benchmark valueCI 1,429 - 1,453

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 #60

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,416
Percentile: 81.4%
Last updated: recent
Eligibility: benchmark_derived_model

81.4% percentile inside its fair comparison set

1,416Raw benchmark valueCI 1,407 - 1,424

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 #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,400
Percentile: 84.2%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_entertainment_and_sports_and_media. Source rank: #63. Votes: 5818. Organization: deepseek. License: MIT.

84.2% percentile inside its fair comparison set

1,400Raw benchmark valueCI 1,391 - 1,408

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 #59

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,431
Percentile: 80.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_legal_and_government. Source rank: #72. Votes: 2174. Organization: deepseek. License: MIT.

80.5% percentile inside its fair comparison set

1,431Raw benchmark valueCI 1,418 - 1,445

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 #43

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,449
Percentile: 87%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_life_and_physical_and_social_science. Source rank: #51. Votes: 4607. Organization: deepseek. License: MIT.

87% percentile inside its fair comparison set

1,449Raw benchmark valueCI 1,439 - 1,458

Text Arena · Industry Mathematical · No Style Control

AR · Professional reasoning · Human

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

Rank #57

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,433
Percentile: 81.8%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_mathematical. Source rank: #67. Votes: 1591. Organization: deepseek. License: MIT.

81.8% percentile inside its fair comparison set

1,433Raw benchmark valueCI 1,418 - 1,449

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 #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,448
Percentile: 83.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_medicine_and_healthcare. Source rank: #58. Votes: 2012. Organization: deepseek. License: MIT.

83.4% percentile inside its fair comparison set

1,448Raw benchmark valueCI 1,434 - 1,462

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 #51

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,449
Percentile: 84.6%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_software_and_it_services. Source rank: #63. Votes: 11498. Organization: deepseek. License: MIT.

84.6% percentile inside its fair comparison set

1,449Raw benchmark valueCI 1,443 - 1,456

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 #41

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,413
Percentile: 87.7%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: industry_writing_and_literature_and_language. Source rank: #52. Votes: 6667. Organization: deepseek. License: MIT.

87.7% percentile inside its fair comparison set

1,413Raw benchmark valueCI 1,405 - 1,421

Data analysis

LB · Professional reasoning · Objective

Structured data manipulation and table reasoning accuracy.

Rank #34

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 68%
Percentile: 69.4%
Last updated: archived
Eligibility: benchmark_derived_model

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

69.4% percentile inside its fair comparison set

68%Raw benchmark value

Overall

LB · Professional reasoning · Objective

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

Rank #45

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 67.2%
Percentile: 59.3%
Last updated: archived
Eligibility: benchmark_derived_model

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

59.3% percentile inside its fair comparison set

67.2%Raw benchmark value

Consecutive events

LB · Professional reasoning · Objective

Objective consecutive events score in LiveBench.

Rank #35

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 59.9%
Percentile: 68.5%
Last updated: archived
Eligibility: benchmark_derived_model

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

68.5% percentile inside its fair comparison set

59.9%Raw benchmark value

Table join

LB · Professional reasoning · Objective

Objective table join score in LiveBench.

Rank #24

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 46.1%
Percentile: 78.7%
Last updated: archived
Eligibility: benchmark_derived_model

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

78.7% percentile inside its fair comparison set

46.1%Raw benchmark value

Table reformat

LB · Professional reasoning · Objective

Objective table reformat score in LiveBench.

Rank #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: LiveBench
Raw value: 98%
Percentile: 74.1%
Last updated: archived
Eligibility: benchmark_derived_model

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

74.1% percentile inside its fair comparison set

98%Raw benchmark value

Search / tool use1 benchmark95.1%

Tau2-Bench Telecom

AA · Search / tool use · Objective

It matters when the model must browse, call tools, and recover useful answers from external systems.

Rank #16 · Source label: DeepSeek V4 Flash (Reasoning, Max Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 95%
Percentile: 95.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `tau2`.

95.1% percentile inside its fair comparison set

95%Raw benchmark value

Long context1 benchmark87.6%

Long Context Reasoning

AA · Long context · Objective

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

Rank #40 · Source label: DeepSeek V4 Flash (Reasoning, High Effort)

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Artificial Analysis
Raw value: 62.7%
Percentile: 87.6%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Artificial Analysis public leaderboard field `lcr`.

87.6% percentile inside its fair comparison set

62.7%Raw benchmark value

Multilingual14 benchmarks80%

Text Arena · Chinese

AR · Multilingual · Human

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

Rank #51

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,478
Percentile: 83.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: chinese. Source rank: #63. Votes: 1386. Organization: deepseek. License: MIT.

83.1% percentile inside its fair comparison set

1,478Raw benchmark valueCI 1,461 - 1,495

Text Arena · French

AR · Multilingual · Human

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

Rank #55

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,450
Percentile: 75%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: french. Source rank: #70. Votes: 1077. Organization: deepseek. License: MIT.

75% percentile inside its fair comparison set

1,450Raw benchmark valueCI 1,429 - 1,470

Text Arena · German

AR · Multilingual · Human

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

Rank #52

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,425
Percentile: 78.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: german. Source rank: #69. Votes: 440. Organization: deepseek. License: MIT.

78.5% percentile inside its fair comparison set

1,425Raw benchmark valueCI 1,397 - 1,454

Text Arena · Japanese

AR · Multilingual · Human

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

Rank #40

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,398
Percentile: 80.8%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: japanese. Source rank: #56. Votes: 285. Organization: deepseek. License: MIT.

80.8% percentile inside its fair comparison set

1,398Raw benchmark valueCI 1,360 - 1,436

Text Arena · Korean

AR · Multilingual · Human

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

Rank #44

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,392
Percentile: 79.3%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: korean. Source rank: #57. Votes: 431. Organization: deepseek. License: MIT.

79.3% percentile inside its fair comparison set

1,392Raw benchmark valueCI 1,362 - 1,422

Text Arena · Russian

AR · Multilingual · Human

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

Rank #49

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,437
Percentile: 83.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: russian. Source rank: #63. Votes: 3180. Organization: deepseek. License: MIT.

83.4% percentile inside its fair comparison set

1,437Raw benchmark valueCI 1,426 - 1,448

Text Arena · Spanish

AR · Multilingual · Human

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

Rank #45

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,437
Percentile: 79.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: spanish. Source rank: #57. Votes: 831. Organization: deepseek. License: MIT.

79.4% percentile inside its fair comparison set

1,437Raw benchmark valueCI 1,415 - 1,458

Text Arena · Chinese · No Style Control

AR · Multilingual · Human

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

Rank #50

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,468
Percentile: 83.4%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: chinese. Source rank: #60. Votes: 1386. Organization: deepseek. License: MIT.

83.4% percentile inside its fair comparison set

1,468Raw benchmark valueCI 1,451 - 1,485

Text Arena · French · No Style Control

AR · Multilingual · Human

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

Rank #57

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,438
Percentile: 74.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: french. Source rank: #69. Votes: 1077. Organization: deepseek. License: MIT.

74.1% percentile inside its fair comparison set

1,438Raw benchmark valueCI 1,418 - 1,459

Text Arena · German · No Style Control

AR · Multilingual · Human

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

Rank #53

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,420
Percentile: 78.1%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: german. Source rank: #66. Votes: 440. Organization: deepseek. License: MIT.

78.1% percentile inside its fair comparison set

1,420Raw benchmark valueCI 1,392 - 1,449

Text Arena · Japanese · No Style Control

AR · Multilingual · Human

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

Rank #43

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,385
Percentile: 79.3%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: japanese. Source rank: #56. Votes: 285. Organization: deepseek. License: MIT.

79.3% percentile inside its fair comparison set

1,385Raw benchmark valueCI 1,347 - 1,422

Text Arena · Korean · No Style Control

AR · Multilingual · Human

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

Rank #53

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,379
Percentile: 75%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: korean. Source rank: #63. Votes: 431. Organization: deepseek. License: MIT.

75% percentile inside its fair comparison set

1,379Raw benchmark valueCI 1,350 - 1,409

Text Arena · Russian · No Style Control

AR · Multilingual · Human

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

Rank #40

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility

Source: Arena
Raw value: 1,431
Percentile: 86.5%
Last updated: recent
Eligibility: benchmark_derived_model

Parsed from Arena leaderboard dataset row `deepseek-v4-flash`. Category: russian. Source rank: #49. Votes: 3180. Organization: deepseek. License: MIT.

86.5% percentile inside its fair comparison set

1,431Raw benchmark valueCI 1,420 - 1,442

Text Arena · Spanish · No Style Control

AR · Multilingual · Human

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

Rank #35

verified runtimeexact aliasBackground only

Raw row drilldownsource row, percentile, last updated, eligibility