Benchmark Leaderboard

Top models by Massive Multitask Language Understanding 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
1Gemini 3.1 Pro Preview98MMLU (accuracy)https://automatio.ai/models/gemini-3-1-pro
2Clef-omni92.7decision-format (Decision Index 0.2.1 internal run)https://huggingface.co/Cloudflare/clef-omni
3GPT-5.592.4MMLU (accuracy)https://tokenmix.ai/blog/gpt-5-5-spud-review-88-swe-bench-2026
4Clef-flash91.8decision-format (Decision Index 0.2.1 internal run)https://huggingface.co/Cloudflare/clef-flash
5Claude Opus 4.691.1MMLU (accuracy)https://automatio.ai/models/claude-opus-4-6
6Clef90.3decision-format (Decision Index 0.2.1 internal run)https://huggingface.co/Cloudflare/clef
7DeepSeek V4 Pro90.15-shothttps://api-docs.deepseek.com/news/news260424
8MiMo-V2.5-Pro89.4—https://huggingface.co/XiaomiMiMo/MiMo-V2.5-Pro
9Claude Sonnet 4.689.3MMLU (accuracy)https://automatio.ai/models/claude-sonnet-4-6
10Claude 3 Opus88.75-shothttps://crfm.stanford.edu/helm/classic/latest/
11Claude 3.5 Sonnet88.75-shothttps://crfm.stanford.edu/helm/classic/latest/
12GPT-4o (05-13)88.75-shothttps://crfm.stanford.edu/helm/classic/latest/
13DeepSeek V4 Flash88.7MMLU EM, DeepSeek-V4-Flash-Base, 5-shothttps://huggingface.co/deepseek-ai/DeepSeek-V4-Flash
14Llama 3.1 405B88.65-shothttps://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard
15Llama 3.1 405B Instruct88.65-shothttps://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard
16DeepSeek V388.55-shothttps://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard
17Qwen2.5-72B-Instruct88.25-shothttps://huggingface.co/spaces/open-llm-leaderboard/open_llm_leaderboard
18Qwen3-Coder-Next87.73From official technical report arXiv:2603 (accuracy)https://arxiv.org/html/2603.00729v1
19Grok-287.55-shothttps://x.ai/blog/grok-2
20GPT-4 Turbo86.55-shothttps://openai.com/index/gpt-4-research/

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.

ModelGPQAHellaSwagHumanEvalMMLU
Gemini 3.1 Pro Preview——
94.0
98.0
Clef-omni
47.4
99.1
—
92.7
GPT-5.5
93.6
—
94.2
92.4
Clef-flash
51.0
98.6
—
91.8
Claude Opus 4.6
91.3
—
95.0
91.1
Clef
48.0
98.2
—
90.3