Benchmarks · /benchmarks/bridgebench-refactoring

Refactoring

Name: Refactoring
Creator: BridgeBench

BridgeBench refactoring benchmark for behavior-preserving code transformations.

Source · BridgeBench
Version · bridgebench snapshot 2026-05-15
Scores · 22

Test details

Verified but agingThis is a rubric-judged signal, so it is more structured than arena taste but still depends on the scoring rubric.

source

BridgeBench

metric

Score (%)

judge

Rubric

direction

higher better

group id

bridgebench_refactoring_2026_05

domain

Coding

What it measures vs what it misses

✓ Measures

Ability to improve or restructure code while preserving expected behavior.

✗ Misses

Large codebase migration ergonomics. Cost and latency.

Why this countsIt tells you whether the model can generate, repair, and reason over code under evaluator pressure rather than marketing examples.Same-test ruleThis percentile only compares models inside the exact benchmark/version group shown here. It is not a universal score.What it missesIt does not fully capture repo-scale iteration, IDE ergonomics, or long debugging loops.

Leaderboard · this benchmark version

#1 · Claude Opus 4.7

BB · May 1, 2026

Source label: openrouter/anthropic/claude-opus-4.7

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 100%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

75.2%

#2 · qwen3.6-plus

BB · May 1, 2026

Source label: openrouter/qwen/qwen3.6-plus

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 95.2%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

74.8%

#3 · Gemini 3.1 Pro Preview

BB · May 1, 2026

Source label: google/gemini-3.1-pro-preview

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 90.5%
Last updated: stale
Eligibility: preview_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

70%

#4 · kimi-k2.6

BB · May 1, 2026

Source label: openrouter/moonshotai/kimi-k2.6

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 85.7%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

69.6%

#5 · Claude Opus 4.6

BB · May 1, 2026

Source label: anthropic/claude-opus-4-6

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 81%
Last updated: stale
Eligibility: historical_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

69.5%

#6 · Claude Sonnet 4.6

BB · May 1, 2026

Source label: anthropic/claude-sonnet-4-6

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 76.2%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

69.4%

#7 · Grok 4.3

BB · May 1, 2026

Source label: openrouter/x-ai/grok-4.3

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 71.4%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

69.1%

#8 · Grok 4.20

BB · May 1, 2026

Source label: x-ai/grok-4.20-reasoning

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 66.7%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

68.1%

#9 · mimo-v2.5-pro

BB · May 1, 2026

Source label: openrouter/xiaomi/mimo-v2.5-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 61.9%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

67.5%

#10 · Qwen3.6 Max Preview

BB · May 1, 2026

Source label: openrouter/qwen/qwen3.6-max-preview

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 57.1%
Last updated: stale
Eligibility: preview_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

66.9%

#11 · GPT-5.5

BB · May 1, 2026

Source label: openai/gpt-5.5

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 52.4%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

65.9%

#12 · GPT-5.4

BB · May 1, 2026

Source label: openai/gpt-5.4

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 47.6%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

63.4%

#13 · GPT-5.4 mini

BB · May 1, 2026

Source label: openai/gpt-5.4-mini

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 42.9%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

62.3%

#14 · glm-5v-turbo

BB · May 1, 2026

Source label: openrouter/z-ai/glm-5v-turbo

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 38.1%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

61%

#15 · glm-5.1

BB · May 1, 2026

Source label: z-ai/glm-5.1

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 33.3%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

60.3%

#16 · GPT-5.4 nano

BB · May 1, 2026

Source label: openai/gpt-5.4-nano

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 28.6%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

56.9%

#17 · mimo-v2.5

BB · May 1, 2026

Source label: openrouter/xiaomi/mimo-v2.5

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 23.8%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

56.5%

#18 · minimax-m2.5

BB · May 1, 2026

Source label: minimax/MiniMax-M2.5

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 19%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

49.2%

#19 · minimax-m2.7

BB · May 1, 2026

Source label: minimax/MiniMax-M2.7

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 14.3%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

49%

#20 · deepseek-v4-pro

BB · May 1, 2026

Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 9.5%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

48.7%

#21 · Kimi K2.5 (Reasoning)

BB · May 1, 2026

Source label: openrouter/moonshotai/kimi-k2.5

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 4.8%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

46.4%

#22 · Nemotron 3 Nano Omni 30B A3B Reasoning

BB · May 1, 2026

Source label: openrouter/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 0%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

38.5%

Benchmarks · /benchmarks/bridgebench-refactoring

Refactoring

BridgeBench refactoring benchmark for behavior-preserving code transformations.

Source · BridgeBench
Version · bridgebench snapshot 2026-05-15
Scores · 22

Test details

Verified but agingThis is a rubric-judged signal, so it is more structured than arena taste but still depends on the scoring rubric.

source

BridgeBench

metric

Score (%)

judge

Rubric

direction

higher better

group id

bridgebench_refactoring_2026_05

domain

Coding

What it measures vs what it misses

✓ Measures

Ability to improve or restructure code while preserving expected behavior.

✗ Misses

Large codebase migration ergonomics. Cost and latency.

Leaderboard · this benchmark version

#1 · Claude Opus 4.7

BB · May 1, 2026

Source label: openrouter/anthropic/claude-opus-4.7

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 100%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

75.2%

#2 · qwen3.6-plus

BB · May 1, 2026

Source label: openrouter/qwen/qwen3.6-plus

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 95.2%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

74.8%

#3 · Gemini 3.1 Pro Preview

BB · May 1, 2026

Source label: google/gemini-3.1-pro-preview

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 90.5%
Last updated: stale
Eligibility: preview_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

70%

#4 · kimi-k2.6

BB · May 1, 2026

Source label: openrouter/moonshotai/kimi-k2.6

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 85.7%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

69.6%

#5 · Claude Opus 4.6

BB · May 1, 2026

Source label: anthropic/claude-opus-4-6

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 81%
Last updated: stale
Eligibility: historical_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

69.5%

#6 · Claude Sonnet 4.6

BB · May 1, 2026

Source label: anthropic/claude-sonnet-4-6

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 76.2%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

69.4%

#7 · Grok 4.3

BB · May 1, 2026

Source label: openrouter/x-ai/grok-4.3

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 71.4%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

69.1%

#8 · Grok 4.20

BB · May 1, 2026

Source label: x-ai/grok-4.20-reasoning

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 66.7%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

68.1%

#9 · mimo-v2.5-pro

BB · May 1, 2026

Source label: openrouter/xiaomi/mimo-v2.5-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 61.9%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

67.5%

#10 · Qwen3.6 Max Preview

BB · May 1, 2026

Source label: openrouter/qwen/qwen3.6-max-preview

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 57.1%
Last updated: stale
Eligibility: preview_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

66.9%

#11 · GPT-5.5

BB · May 1, 2026

Source label: openai/gpt-5.5

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 52.4%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

65.9%

#12 · GPT-5.4

BB · May 1, 2026

Source label: openai/gpt-5.4

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 47.6%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

63.4%

#13 · GPT-5.4 mini

BB · May 1, 2026

Source label: openai/gpt-5.4-mini

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 42.9%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

62.3%

#14 · glm-5v-turbo

BB · May 1, 2026

Source label: openrouter/z-ai/glm-5v-turbo

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 38.1%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

61%

#15 · glm-5.1

BB · May 1, 2026

Source label: z-ai/glm-5.1

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 33.3%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

60.3%

#16 · GPT-5.4 nano

BB · May 1, 2026

Source label: openai/gpt-5.4-nano

verified runtimeexact alias

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 28.6%
Last updated: stale
Eligibility: headline eligible
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

56.9%

#17 · mimo-v2.5

BB · May 1, 2026

Source label: openrouter/xiaomi/mimo-v2.5

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 23.8%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

56.5%

#18 · minimax-m2.5

BB · May 1, 2026

Source label: minimax/MiniMax-M2.5

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 19%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

49.2%

#19 · minimax-m2.7

BB · May 1, 2026

Source label: minimax/MiniMax-M2.7

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 14.3%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

49%

#20 · deepseek-v4-pro

BB · May 1, 2026

Source label: openrouter/deepseek/deepseek-v4-pro

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 9.5%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

48.7%

#21 · Kimi K2.5 (Reasoning)

BB · May 1, 2026

Source label: openrouter/moonshotai/kimi-k2.5

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 4.8%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

46.4%

#22 · Nemotron 3 Nano Omni 30B A3B Reasoning

BB · May 1, 2026

Source label: openrouter/nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free

verified runtimeexact aliasBackground only

Raw row drilldownsource, percentile, eligibility

Source URL: https://www.bridgebench.ai/refactoring
Percentile: 0%
Last updated: stale
Eligibility: benchmark_derived_model
Identity: provider alias (0.92)

Parsed from the BridgeBench page for bridgebench-refactoring.

38.5%

Refactoring

Test details

What it measures vs what it misses

✓ Measures

✗ Misses

Leaderboard · this benchmark version

Loading benchmark evidence.

Refactoring

Test details

What it measures vs what it misses

✓ Measures

✗ Misses

Leaderboard · this benchmark version