Benchmark Almanac

PowerCodeBench

Llama 3.1-405B 68.7% leads since May 2026
Not in model cards

Models translate operators' requests into Python programs for pandapower. The evaluator runs each program and compares its numerical or Boolean output with the reference answer.

Grid Studio scored 352 of the 2,000 tasks. It restored missing specifications from the source records for 156 of those questions.

Released
May 2026
Kind
Scientific code
Level
University
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Code
Interaction
Single turn
Environment
Code sandbox
Refresh
One fixed set
Run by
Anyone
Occupations
Production
Strongest model tested
GPT-5.4 Mini #74 on the capability index
Last model added
April 2026
Scores refreshed
November 20th, 2026
15
Entries on its own leaderboard, 12 models we track

Llama 3.1-405B leads with 68.7% among models released by April 2026 Llama 3.1-405B leads with 68.7% among models released by April 2026

Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Llama 3.1-405B C_FDRS 68.7%
  2. 2 Qwen3-Coder-480B-A35B X_FDRS 67.3%
  3. 3 Claude Haiku 4.5 C_FDRS 63.8%
  4. 4 GPT-5.4 Mini C_FDRS 62.1%
  5. 5 Qwen3-Coder-Next X_FDRS 60.9%
  6. 6 gpt-oss-120b C_FDRS 60.3%
  7. 7 Llama 3.1-70B C_FDRS 57.5%
  8. 8 DeepSeek-V4-Flash C_FDRS 55.9%
  9. 9 Qwen2.5-Coder-32B-Instruct X_FD 54.3%
  10. 10 Gemini 2.5 Flash Sep 2025, C_FDRS 52%
  11. 11 Qwen2.5-Coder-14B-Instruct X_FDRS 51.1%
  12. 12 Llama 3.1-8B X_FDRS 36.5%
  13. 13 Qwen2.5-Coder-7B-Instruct C_FDRS 32.8%
  14. 14 Qwen2.5-Coder-1.5B-Instruct C_FX 9.90%
  15. 15 Qwen2.5-Coder-0.5B-Instruct X_FD 0.60%

Source: Official PowerCodeBench leaderboard

#EntryScoreSettingDateSource
1 Llama 3.1-405B 68.7% C_FDRS Jul 2024 Official leaderboard
2 Qwen3-Coder-480B-A35B 67.3% X_FDRS Jul 2025 Official leaderboard
3 Claude Haiku 4.5 63.8% C_FDRS Oct 2025 Official leaderboard
4 GPT-5.4 Mini 62.1% C_FDRS Mar 2026 Official leaderboard
5 Qwen3-Coder-Next 60.9% X_FDRS Feb 2026 Official leaderboard
6 gpt-oss-120b 60.3% C_FDRS Aug 2025 Official leaderboard
7 Llama 3.1-70B 57.5% C_FDRS Jul 2024 Official leaderboard
8 DeepSeek-V4-Flash 55.9% C_FDRS Apr 2026 Official leaderboard
9 Qwen2.5-Coder-32B-Instruct 54.3% X_FD Nov 2024 Official leaderboard
10 Gemini 2.5 Flash Sep 2025 52% C_FDRS Sep 2025 Official leaderboard
11 Qwen2.5-Coder-14B-Instruct 51.1% X_FDRS Nov 2024 Official leaderboard
12 Llama 3.1-8B 36.5% X_FDRS Jul 2024 Official leaderboard
13 Qwen2.5-Coder-7B-Instruct 32.8% C_FDRS Official leaderboard
14 Qwen2.5-Coder-1.5B-Instruct 9.90% C_FX Official leaderboard
15 Qwen2.5-Coder-0.5B-Instruct 0.60% X_FD Official leaderboard
#ModelScoreSettingDateSource
1 Llama 3.1-405B 68.7% C_FDRS Jul 2024 Official leaderboard
2 Qwen3-Coder-480B-A35B 67.3% X_FDRS Jul 2025 Official leaderboard
3 Claude Haiku 4.5 63.8% C_FDRS Oct 2025 Official leaderboard
4 GPT-5.4 Mini 62.1% C_FDRS Mar 2026 Official leaderboard
5 Qwen3-Coder-Next 60.9% X_FDRS Feb 2026 Official leaderboard
6 gpt-oss-120b 60.3% C_FDRS Aug 2025 Official leaderboard
7 Llama 3.1-70B 57.5% C_FDRS Jul 2024 Official leaderboard
8 DeepSeek-V4-Flash 55.9% C_FDRS Apr 2026 Official leaderboard
9 Qwen2.5-Coder-32B-Instruct 54.3% X_FD Nov 2024 Official leaderboard
10 Gemini 2.5 Flash Sep 2025 52% C_FDRS Sep 2025 Official leaderboard
11 Qwen2.5-Coder-14B-Instruct 51.1% X_FDRS Nov 2024 Official leaderboard
12 Llama 3.1-8B 36.5% X_FDRS Jul 2024 Official leaderboard

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Llama 3.1-405B
B A A_FX A_FDRS X X_FX X_FDRS
68.7%
Qwen3-Coder-480B-A35B
A B A_FDR C C_FDR
67.3%
Claude Haiku 4.5
A C C_FDR
63.8%
GPT-5.4 Mini
A C C_FDR
62.1%
Qwen3-Coder-Next
A A_FDR C C_FX X_FDR
60.9%
gpt-oss-120b
B A_FX C X_FX
60.3%
Llama 3.1-70B
A A_FX A_FDR C C_FX X_FDRS
57.5%
DeepSeek-V4-Flash
A C C_FDR C_FDRS
55.9%
Qwen2.5-Coder-32B-Instruct
A A_FX C C_FX
54.3%
Gemini 2.5 Flash
A C C_FX
52%
Qwen2.5-Coder-14B-Instruct
A B C C_FX C_FDRS
51.1%
Llama 3.1-8B
A C X_FX X_FDR
36.5%

Source: Official PowerCodeBench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Llama 3.1-405B C_FDRS 68.7%
Qwen3-Coder-480B-A35B X_FDRS 67.3%
Qwen3-Coder-480B-A35B X_FDR 66.6%
Llama 3.1-405B C_FD 66.5%
Qwen3-Coder-480B-A35B X_FX 66.3%
Qwen3-Coder-480B-A35B C_FDRS 65.9%
Qwen3-Coder-480B-A35B X_FD 65.5%
Qwen3-Coder-480B-A35B C_FD 65.5%
Llama 3.1-405B X_FDRS 65.3%
Llama 3.1-405B C_FDR 64.8%
Llama 3.1-405B C_FX 64.8%
Qwen3-Coder-480B-A35B C_FX 63.8%
Qwen3-Coder-480B-A35B C_FDR 63.8%
Claude Haiku 4.5 C_FDRS 63.8%
GPT-5.4 Mini C_FDRS 62.1%
GPT-5.4 Mini C_FX 61.3%
GPT-5.4 Mini C_FDR 61.1%
Claude Haiku 4.5 C_FDR 61.1%
Claude Haiku 4.5 C_FX 61.1%
Qwen3-Coder-Next X_FDRS 60.9%
Llama 3.1-405B X_FDR 60.6%
Qwen3-Coder-Next X_FDR 60.6%
Llama 3.1-405B X_FD 60.3%
gpt-oss-120b C_FDRS 60.3%
Qwen3-Coder-Next X_FD 60.3%
gpt-oss-120b C_FD 59.8%
Qwen3-Coder-Next C_FDRS 59.5%
Qwen3-Coder-480B-A35B X 59.5%
Llama 3.1-405B X_FX 59.3%
Qwen3-Coder-Next X_FX 58.8%
gpt-oss-120b C_FX 58.4%
gpt-oss-120b C_FDR 58.4%
Qwen3-Coder-Next C_FDR 58%
Qwen3-Coder-Next C_FD 57.9%
Llama 3.1-70B C_FDRS 57.5%
Llama 3.1-70B X_FDRS 57%
gpt-oss-120b X_FD 56.8%
gpt-oss-120b X_FDRS 56.7%
Qwen3-Coder-480B-A35B C 56.6%
gpt-oss-120b X_FDR 56.5%
gpt-oss-120b X_FX 56.5%
DeepSeek-V4-Flash C_FDRS 55.9%
Qwen3-Coder-Next C_FX 55%
Qwen2.5-Coder-32B-Instruct X_FD 54.3%
Qwen2.5-Coder-32B-Instruct X_FDR 54.3%
Qwen2.5-Coder-32B-Instruct X_FDRS 54.1%
Llama 3.1-70B X_FDR 52.8%
Qwen2.5-Coder-32B-Instruct C_FDR 52.8%
Qwen2.5-Coder-32B-Instruct C_FDRS 52.8%
Llama 3.1-70B X_FD 52.2%
Qwen2.5-Coder-32B-Instruct C_FD 52.2%
Qwen3-Coder-Next X 52.1%
Gemini 2.5 Flash Sep 2025 C_FDRS 52%
Llama 3.1-70B X_FX 51.8%
Llama 3.1-405B C 51.5%
Llama 3.1-70B C_FDR 51.5%
Qwen2.5-Coder-32B-Instruct X_FX 51.3%
Qwen2.5-Coder-14B-Instruct X_FDRS 51.1%
Llama 3.1-70B C_FD 51%
Qwen2.5-Coder-32B-Instruct C_FX 50.8%
GPT-5.4 Mini C 50.5%
Qwen2.5-Coder-14B-Instruct X_FDR 50.1%
Gemini 2.5 Flash Sep 2025 C_FDR 49.4%
Llama 3.1-70B C_FX 49%
gpt-oss-120b X 49%
Gemini 2.5 Flash Sep 2025 C_FX 48.6%
Qwen2.5-Coder-14B-Instruct C_FDRS 48.6%
Qwen2.5-Coder-14B-Instruct X_FD 48%
Llama 3.1-405B X 47.5%
DeepSeek-V4-Flash C_FDR 47.3%
DeepSeek-V4-Flash C_FX 47.3%
Qwen3-Coder-Next C 47.2%
Qwen2.5-Coder-14B-Instruct X_FX 46.5%
Qwen2.5-Coder-14B-Instruct C_FD 46.5%
Qwen2.5-Coder-14B-Instruct C_FDR 46.1%
Qwen2.5-Coder-32B-Instruct X 46%
gpt-oss-120b C 45.6%
Qwen2.5-Coder-32B-Instruct C 42.7%
Qwen2.5-Coder-14B-Instruct C_FX 42.5%
Llama 3.1-70B X 41.6%
DeepSeek-V4-Flash C 40.6%
Gemini 2.5 Flash Sep 2025 C 40%
Qwen2.5-Coder-14B-Instruct X 39.4%
Llama 3.1-70B C 38.1%
Claude Haiku 4.5 C 36.9%
Llama 3.1-405B A_FDRS 36.5%
Llama 3.1-8B X_FDRS 36.5%
Llama 3.1-8B C_FDRS 36.3%
Llama 3.1-8B C_FDR 34.4%
Llama 3.1-8B X_FDR 34.2%
Qwen2.5-Coder-14B-Instruct C 34.2%
Llama 3.1-405B A_FDR 33%
Llama 3.1-8B C_FD 31.1%
Qwen3-Coder-480B-A35B A_FDRS 31.1%
Llama 3.1-8B X_FD 30.6%
Llama 3.1-405B A_FD 29.3%
Llama 3.1-8B C_FX 28.9%
Llama 3.1-405B A_FX 28.6%
Llama 3.1-8B X_FX 28.5%
Qwen3-Coder-480B-A35B A_FDR 28.5%
Qwen3-Coder-480B-A35B A_FD 27.1%
gpt-oss-120b A_FDRS 26.6%
Qwen3-Coder-480B-A35B A_FX 26.5%
gpt-oss-120b A_FDR 25.3%
GPT-5.4 Mini A 24.3%
gpt-oss-120b A_FD 23.9%
Llama 3.1-70B A_FDRS 23.1%
gpt-oss-120b A_FX 22.9%
Qwen3-Coder-480B-A35B B 21.6%
Llama 3.1-70B A_FDR 20.9%
Llama 3.1-8B X 20.4%
Llama 3.1-8B C 20.1%
Llama 3.1-70B A_FD 17.8%
Qwen2.5-Coder-32B-Instruct A_FDRS 15.8%
Qwen2.5-Coder-14B-Instruct B 15.6%
Qwen3-Coder-Next B 15.6%
Qwen2.5-Coder-32B-Instruct A_FDR 15.4%
Qwen3-Coder-480B-A35B A 15.1%
Llama 3.1-70B A_FX 15%
Qwen2.5-Coder-32B-Instruct B 14.9%
gpt-oss-120b A 14.1%
Llama 3.1-405B A 13.4%
Qwen2.5-Coder-32B-Instruct A_FD 12.2%
Qwen2.5-Coder-32B-Instruct A_FX 11.2%
DeepSeek-V4-Flash A 11.1%
Qwen3-Coder-Next A_FDRS 10.5%
Claude Haiku 4.5 A 10.3%
Qwen3-Coder-Next A_FDR 10.3%
gpt-oss-120b B 9.65%
Qwen3-Coder-Next A_FD 7.95%
Qwen2.5-Coder-14B-Instruct A_FDRS 6.95%
Qwen3-Coder-Next A_FX 6.95%
Qwen2.5-Coder-14B-Instruct A_FDR 6.25%
Llama 3.1-70B B 5.20%
Llama 3.1-70B A 5.00%
Llama 3.1-405B B 4.20%
Gemini 2.5 Flash Sep 2025 A 4.10%
Qwen3-Coder-Next A 4.00%
Qwen2.5-Coder-14B-Instruct A_FX 3.85%
Qwen2.5-Coder-32B-Instruct A 3.80%
Qwen2.5-Coder-14B-Instruct A_FD 3.45%
Llama 3.1-8B A_FDRS 2.85%
Llama 3.1-8B A_FDR 2.60%
Qwen2.5-Coder-14B-Instruct A 1.60%
Llama 3.1-8B B 1.30%
Llama 3.1-8B A_FX 1.25%
Llama 3.1-8B A_FD 1.00%
Llama 3.1-8B A 0.50%

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
Llama 3.1-405B
B A A_FX A_FDRS X X_FX X_FDRS
68.7%
Qwen3-Coder-480B-A35B
A B A_FDR C C_FDR
67.3%
Claude Haiku 4.5
A C C_FDR
63.8%
GPT-5.4 Mini
A C C_FDR
62.1%
Qwen3-Coder-Next
A A_FDR C C_FX X_FDR
60.9%
gpt-oss-120b
B A_FX C X_FX
60.3%
Llama 3.1-70B
A A_FX A_FDR C C_FX X_FDRS
57.5%
DeepSeek-V4-Flash
A C C_FDR C_FDRS
55.9%
Qwen2.5-Coder-32B-Instruct
A A_FX C C_FX
54.3%
Gemini 2.5 Flash
A C C_FX
52%
Qwen2.5-Coder-14B-Instruct
A B C C_FX C_FDRS
51.1%
Llama 3.1-8B
A C X_FX X_FDR
36.5%

Source: Official PowerCodeBench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
Llama 3.1-405B C_FDRS 68.7%
Qwen3-Coder-480B-A35B X_FDRS 67.3%
Qwen3-Coder-480B-A35B X_FDR 66.6%
Llama 3.1-405B C_FD 66.5%
Qwen3-Coder-480B-A35B X_FX 66.3%
Qwen3-Coder-480B-A35B C_FDRS 65.9%
Qwen3-Coder-480B-A35B X_FD 65.5%
Qwen3-Coder-480B-A35B C_FD 65.5%
Llama 3.1-405B X_FDRS 65.3%
Llama 3.1-405B C_FDR 64.8%
Llama 3.1-405B C_FX 64.8%
Qwen3-Coder-480B-A35B C_FX 63.8%
Qwen3-Coder-480B-A35B C_FDR 63.8%
Claude Haiku 4.5 C_FDRS 63.8%
GPT-5.4 Mini C_FDRS 62.1%
GPT-5.4 Mini C_FX 61.3%
GPT-5.4 Mini C_FDR 61.1%
Claude Haiku 4.5 C_FDR 61.1%
Claude Haiku 4.5 C_FX 61.1%
Qwen3-Coder-Next X_FDRS 60.9%
Llama 3.1-405B X_FDR 60.6%
Qwen3-Coder-Next X_FDR 60.6%
Llama 3.1-405B X_FD 60.3%
gpt-oss-120b C_FDRS 60.3%
Qwen3-Coder-Next X_FD 60.3%
gpt-oss-120b C_FD 59.8%
Qwen3-Coder-Next C_FDRS 59.5%
Qwen3-Coder-480B-A35B X 59.5%
Llama 3.1-405B X_FX 59.3%
Qwen3-Coder-Next X_FX 58.8%
gpt-oss-120b C_FX 58.4%
gpt-oss-120b C_FDR 58.4%
Qwen3-Coder-Next C_FDR 58%
Qwen3-Coder-Next C_FD 57.9%
Llama 3.1-70B C_FDRS 57.5%
Llama 3.1-70B X_FDRS 57%
gpt-oss-120b X_FD 56.8%
gpt-oss-120b X_FDRS 56.7%
Qwen3-Coder-480B-A35B C 56.6%
gpt-oss-120b X_FDR 56.5%
gpt-oss-120b X_FX 56.5%
DeepSeek-V4-Flash C_FDRS 55.9%
Qwen3-Coder-Next C_FX 55%
Qwen2.5-Coder-32B-Instruct X_FD 54.3%
Qwen2.5-Coder-32B-Instruct X_FDR 54.3%
Qwen2.5-Coder-32B-Instruct X_FDRS 54.1%
Llama 3.1-70B X_FDR 52.8%
Qwen2.5-Coder-32B-Instruct C_FDR 52.8%
Qwen2.5-Coder-32B-Instruct C_FDRS 52.8%
Llama 3.1-70B X_FD 52.2%
Qwen2.5-Coder-32B-Instruct C_FD 52.2%
Qwen3-Coder-Next X 52.1%
Gemini 2.5 Flash Sep 2025 C_FDRS 52%
Llama 3.1-70B X_FX 51.8%
Llama 3.1-405B C 51.5%
Llama 3.1-70B C_FDR 51.5%
Qwen2.5-Coder-32B-Instruct X_FX 51.3%
Qwen2.5-Coder-14B-Instruct X_FDRS 51.1%
Llama 3.1-70B C_FD 51%
Qwen2.5-Coder-32B-Instruct C_FX 50.8%
GPT-5.4 Mini C 50.5%
Qwen2.5-Coder-14B-Instruct X_FDR 50.1%
Gemini 2.5 Flash Sep 2025 C_FDR 49.4%
Llama 3.1-70B C_FX 49%
gpt-oss-120b X 49%
Gemini 2.5 Flash Sep 2025 C_FX 48.6%
Qwen2.5-Coder-14B-Instruct C_FDRS 48.6%
Qwen2.5-Coder-14B-Instruct X_FD 48%
Llama 3.1-405B X 47.5%
DeepSeek-V4-Flash C_FDR 47.3%
DeepSeek-V4-Flash C_FX 47.3%
Qwen3-Coder-Next C 47.2%
Qwen2.5-Coder-14B-Instruct X_FX 46.5%
Qwen2.5-Coder-14B-Instruct C_FD 46.5%
Qwen2.5-Coder-14B-Instruct C_FDR 46.1%
Qwen2.5-Coder-32B-Instruct X 46%
gpt-oss-120b C 45.6%
Qwen2.5-Coder-32B-Instruct C 42.7%
Qwen2.5-Coder-14B-Instruct C_FX 42.5%
Llama 3.1-70B X 41.6%
DeepSeek-V4-Flash C 40.6%
Gemini 2.5 Flash Sep 2025 C 40%
Qwen2.5-Coder-14B-Instruct X 39.4%
Llama 3.1-70B C 38.1%
Claude Haiku 4.5 C 36.9%
Llama 3.1-405B A_FDRS 36.5%
Llama 3.1-8B X_FDRS 36.5%
Llama 3.1-8B C_FDRS 36.3%
Llama 3.1-8B C_FDR 34.4%
Llama 3.1-8B X_FDR 34.2%
Qwen2.5-Coder-14B-Instruct C 34.2%
Llama 3.1-405B A_FDR 33%
Llama 3.1-8B C_FD 31.1%
Qwen3-Coder-480B-A35B A_FDRS 31.1%
Llama 3.1-8B X_FD 30.6%
Llama 3.1-405B A_FD 29.3%
Llama 3.1-8B C_FX 28.9%
Llama 3.1-405B A_FX 28.6%
Llama 3.1-8B X_FX 28.5%
Qwen3-Coder-480B-A35B A_FDR 28.5%
Qwen3-Coder-480B-A35B A_FD 27.1%
gpt-oss-120b A_FDRS 26.6%
Qwen3-Coder-480B-A35B A_FX 26.5%
gpt-oss-120b A_FDR 25.3%
GPT-5.4 Mini A 24.3%
gpt-oss-120b A_FD 23.9%
Llama 3.1-70B A_FDRS 23.1%
gpt-oss-120b A_FX 22.9%
Qwen3-Coder-480B-A35B B 21.6%
Llama 3.1-70B A_FDR 20.9%
Llama 3.1-8B X 20.4%
Llama 3.1-8B C 20.1%
Llama 3.1-70B A_FD 17.8%
Qwen2.5-Coder-32B-Instruct A_FDRS 15.8%
Qwen2.5-Coder-14B-Instruct B 15.6%
Qwen3-Coder-Next B 15.6%
Qwen2.5-Coder-32B-Instruct A_FDR 15.4%
Qwen3-Coder-480B-A35B A 15.1%
Llama 3.1-70B A_FX 15%
Qwen2.5-Coder-32B-Instruct B 14.9%
gpt-oss-120b A 14.1%
Llama 3.1-405B A 13.4%
Qwen2.5-Coder-32B-Instruct A_FD 12.2%
Qwen2.5-Coder-32B-Instruct A_FX 11.2%
DeepSeek-V4-Flash A 11.1%
Qwen3-Coder-Next A_FDRS 10.5%
Claude Haiku 4.5 A 10.3%
Qwen3-Coder-Next A_FDR 10.3%
gpt-oss-120b B 9.65%
Qwen3-Coder-Next A_FD 7.95%
Qwen2.5-Coder-14B-Instruct A_FDRS 6.95%
Qwen3-Coder-Next A_FX 6.95%
Qwen2.5-Coder-14B-Instruct A_FDR 6.25%
Llama 3.1-70B B 5.20%
Llama 3.1-70B A 5.00%
Llama 3.1-405B B 4.20%
Gemini 2.5 Flash Sep 2025 A 4.10%
Qwen3-Coder-Next A 4.00%
Qwen2.5-Coder-14B-Instruct A_FX 3.85%
Qwen2.5-Coder-32B-Instruct A 3.80%
Qwen2.5-Coder-14B-Instruct A_FD 3.45%
Llama 3.1-8B A_FDRS 2.85%
Llama 3.1-8B A_FDR 2.60%
Qwen2.5-Coder-14B-Instruct A 1.60%
Llama 3.1-8B B 1.30%
Llama 3.1-8B A_FX 1.25%
Llama 3.1-8B A_FD 1.00%
Llama 3.1-8B A 0.50%

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

Meta
Own leaderboard Not tested since May 2026

Source: Official PowerCodeBench leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard
0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Llama 3.1-405B 0% 20% 40% 60% 80% 100% 2025 2026 Near the ceiling (85%) Llama 3.1-405B

Source: Official PowerCodeBench leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Coding

All 523 in Coding
  • SWE-bench Verified In use 500 human-validated GitHub issues from 12 Python repos. Agent writes a patch that must pass hidden tests. Claude Opus 4.7 83.5%
  • LiveCodeBench v6 In use Competitive programming problems collected after model cutoffs from LeetCode, AtCoder, Codeforces. Scored pass@1 on hidden tests. DeepSeek-V4-Pro 93.5%
  • SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Fable 5.2 96.6%
  • DeepSWE In use 113 original long-horizon engineering tasks across 91 repos and five languages, written from scratch. Hand-written verifiers accept alternative solutions. GPT-6 Astra 74.1%
  • SciCode In use Scientist-curated research coding problems split into subproblems across natural sciences. Code checked against test cases. Seed2.1 Pro 59.8%
  • Codeforces rating In use Model solves recent Codeforces contest problems. An Elo-style rating is computed from simulated contest performance. DeepSeek V4.1 Flash 3,471

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for DeepSeek-V4-Flash. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (15 entries).
  • Pass Models scored 12 models we track, plus 3 other entries on its leaderboard.
  • Warning Newest score The newest came 7 months ago, in April 2026, for DeepSeek-V4-Flash.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.4 Mini (#74).
  • Pass Headroom The best is 68.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on github.com: 148 rows for models we track and 3 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check