Wikibench

GPQA

reasoning official site →

Multilingual: no

A challenging dataset of 448 multiple-choice questions written by domain experts in biology, physics, and chemistry. Questions are Google-proof and extremely difficult, with PhD experts reaching 65% accuracy.

Methodology

Imported from llm-stats public benchmark metadata. Modality: text. Max score: 1. Categories: biology, chemistry, general, physics, reasoning. Language: en. Verified by llm-stats: no.

Leaderboard

Showing 20 of 207 results

  1. Claude Mythos Preview self-reported llm-stats
    94.6%
  2. Gemini 3.1 Pro self-reported llm-stats
    94.3%
  3. Claude Opus 4.7 self-reported llm-stats
    94.2%
  4. Claude Opus 4.8 self-reported llm-stats
    93.6%
  5. GPT-5.5 self-reported llm-stats
    93.6%
  6. GPT-5.2 Pro self-reported llm-stats
    93.2%
  7. GPT-5.4 self-reported llm-stats
    92.8%
  8. GPT-5.2 self-reported llm-stats
    92.4%
  9. Qwen3.7 Max self-reported llm-stats
    92.4%
  10. Gemini 3 Pro self-reported llm-stats
    91.9%
  11. Claude Opus 4.6 self-reported llm-stats
    91.3%
  12. Kimi K2.6 self-reported llm-stats
    90.5%
  13. Gemini 3 Flash self-reported llm-stats
    90.4%
  14. Qwen3.6 Plus self-reported llm-stats
    90.4%
  15. DeepSeek-V4-Pro-Max self-reported llm-stats
    90.1%
  16. Claude Sonnet 4.6 self-reported llm-stats
    89.9%
  17. Muse Spark self-reported llm-stats
    89.5%
  18. Seed 2.0 Pro self-reported llm-stats
    88.9%
  19. Grok-4 Heavy self-reported llm-stats
    88.4%
  20. Qwen3.5-397B-A17B self-reported llm-stats
    88.4%