LLM Reference

Benchmark Leaderboard

Top models by HumanEval score

Sorts are only valid inside this benchmark. A score gap of this benchmark does not map to the same gap on another benchmark, so validate any short-list by checking peer rows and model context/price fit.

#ModelScoreVersionSource
1Claude Sonnet 4.698HumanEval (pass@1)https://automatio.ai/models/claude-sonnet-4-6
2o396.72025-04https://openai.com/index/introducing-o3-and-o4-mini/
3Claude Opus 4.695HumanEval (pass@1)https://automatio.ai/models/claude-opus-4-6
4Grok-394.5https://x.ai/blog/grok-3
5GPT-5.594.2https://openai.com/index/introducing-gpt-5-5/
6Gemini 3.1 Pro Preview94HumanEval (pass@1)https://automatio.ai/models/gemini-3-1-pro
7Gemini 2.5 Pro93.12025-03https://storage.googleapis.com/deepmind-media/gemini/gemini_v2_5_report.pdf
8Claude 3.7 Sonnet932025-02https://www.anthropic.com/news/claude-3-7-sonnet
9GPT-4.192.92025-04https://openai.com/index/gpt-4-1/
10Qwen2.5-Coder-32B-Instruct92.72024-11https://qwenlm.github.io/blog/qwen2.5-coder-family/
11Qwen3-235B-A22B92.72025-04https://qwenlm.github.io/blog/qwen3/
12Claude 3.5 Sonnet92pass@1https://crfm.stanford.edu/helm/classic/latest/
13Kimi K2.692https://moonshotai.github.io/Kimi-K2/
14Mistral Large 3 675B Instruct92https://mistral.ai/news/mistral-large-3/
15Gemini 3.5 Flash92https://o-mega.ai/articles/gemini-3-5-flash-benchmarks-cost-and-guide
16OLMo 3 32B Think91.4HumanEval (pass@1)https://huggingface.co/allenai/Olmo-3-32B-Think
17GPT-4o (05-13)90.2pass@1https://crfm.stanford.edu/helm/classic/latest/
18Gemini 2.5 Flash90.12025-05https://storage.googleapis.com/deepmind-media/gemini/gemini_v2_5_report.pdf
19DeepSeek R189.92025-01https://arxiv.org/abs/2501.12948
20Granite 4.1 30B89.63pass@1, instruct model (pass@1)https://huggingface.co/blog/ibm-granite/granite-4-1

Interpretation

Trust when:

  • Use top rows for shortlisting, then validate provider coverage and feature fit before selecting a model.
  • Cross-benchmark consistency is a stronger signal than one-off score dominance.

Don't trust when:

  • Don't compare raw score gaps across different benchmark scales as equivalent.
  • Missing rows in the heatmap can hide benchmark blind spots.
  • No trend data means this is a snapshot; expect drift in fresh refresh cycles.

Cross-benchmark heatmap

Compare top models across selected benchmark families to spot strengths and weak spots. On narrow screens, scroll horizontally to inspect every benchmark.

ModelMMLUGPQAHumanEvalHellaSwag
Claude Sonnet 4.6
89.3
89.9
98.0
o3
87.7
96.7
Claude Opus 4.6
91.1
91.3
95.0
Grok-3
84.6
94.5
GPT-5.5
92.4
93.6
94.2
Gemini 3.1 Pro Preview
98.0
94.3
94.0