LLM Reference

Benchmark Leaderboard

Top models by Google-Proof Q&A 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 Mythos Preview94.6diamondhttps://epoch.ai/benchmarks/gpqa-diamond; https://intuitionlabs.ai/articles/gpqa-diamond-ai-benchmark
2GPT-5.6 Sol94.6GPQA Diamond; max reasoninghttps://openai.com/index/gpt-5-6/
3Gemini 3.1 Pro Preview94.3diamondhttps://artificialanalysis.ai/leaderboards/models
4Claude Opus 4.794.2diamondhttps://www.anthropic.com/news/claude-opus-4-7
5GPT-5.593.6diamondhttps://openai.com/index/introducing-gpt-5-5/
6GPT-5.5 Pro93.6GPQA Diamondhttps://codingfleet.com/blog/claude-opus-4-8-vs-gpt-5-5-comparison/
7Claude Opus 4.893.6GPQA Diamondhttps://www.anthropic.com/news/claude-opus-4-8
8Kimi K393.5GPQA-Diamond; max reasoning efforthttps://www.kimi.com/blog/kimi-k3
9MiniMax M392.9MiniMax M3 GPQA 92 (accuracy%)https://venturebeat.com/technology/minimax-m3-debuts-eclipsing-gpt-5-5-and-gemini-3-1-pro-on-key-benchmark-performance-for-just-5-10-of-the-cost
10Qwen3.7-Max92.4diamondhttps://www.datacamp.com/blog/qwen3-7-max
11Gemini 3.5 Flash92.2GPQA Diamond (accuracy)https://www.nxcode.io/resources/news/gemini-3-5-flash-complete-guide-benchmarks-pricing-api-2026
12GPT-5.492diamondhttps://pricepertoken.com/leaderboards/benchmark/gpqa
13Gemini 3 Pro91.9https://deepmind.google/technologies/gemini/pro/
14Qwen3.6-Max91.8https://qwenlm.github.io/blog/qwen3.6/
15Claude Opus 4.691.3diamondhttps://www.anthropic.com/claude/opus
16GLM-5.291.2diamondhttps://huggingface.co/zai-org/GLM-5.2
17Kimi K2.690.5GPQA Diamond (accuracy)https://huggingface.co/moonshotai/Kimi-K2.6
18Gemini 3 Flash90.4https://deepmind.google/technologies/gemini/flash/
19Qwen3.6-Plus90.4llm-stats shows 0 (accuracy%)https://llm-stats.com/benchmarks/gpqa
20DeepSeek V4 Pro90.1diamondhttps://api-docs.deepseek.com/news/news260424

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 Mythos Preview
94.6
GPT-5.6 Sol
94.6
Gemini 3.1 Pro Preview
98.0
94.3
94.0
Claude Opus 4.7
94.2
GPT-5.5
92.4
93.6
94.2
GPT-5.5 Pro
93.6