Vibeleaderboard
← Benchmark guide

Model results

GPQA Diamond

Choose a shared evaluation protocol to compare model scores. Versions, sources and configurations stay separate.

Atlas snapshot captured 10/19/2026, 10:20:28 AM UTC. This is an import timestamp, not the evaluation date. Coverage may be incomplete.

Reviewed recommendation and limits

Use as a compact reference for expert-written science questions, alongside evidence that matches the scientific work you intend to do.

Original GPQA Diamond subset. Preserve prompt, answer extraction, tools, sampling and aggregation. Keep main, extended and third-party corrected subsets separate.. Imported groups below may use different protocols; this review does not certify each run. The static 198-question subset has limited statistical power and exposure risk. Its expert filtering does not make it representative of scientific practice. A Google-proof collection process does not prove present-day contamination resistance.

Publisher results GPQA paper
← All evaluation groups

GPQA Diamond

Source
Epoch AI Benchmarking Hub
Version
Not reported
Measurement date
Not reported
Metric
% · Higher is better
Evaluation scope
Epoch AI run · task v1 · no reasoning · result date not supplied

Exact score ties share a position. Positions do not establish statistical significance. Missing metadata is shown as unreported; matching imported metadata is not independent verification of the protocol.

  1. #1Qwen3.5 397B A17B

    qwen/qwen3.5-397b-a17b-20260216

    86.4%

    View modelEvidence independent

  2. #2Claude Opus 4.8

    anthropic/claude-opus-4-8

    85.4%

    View modelEvidence independent

  3. #3Qwen3.6 27B

    qwen/qwen3.6-27b-20260422

    84.8%

    View modelEvidence independent

  4. #3Qwen3.6 35B A3B

    qwen/qwen3.6-35b-a3b-20260415

    84.8%

    View modelEvidence independent

  5. #5GPT-5.6 Sol

    openai/gpt-5.6-sol-20260709

    82.8%

    View modelEvidence independent

  6. #6Qwen3.7 Plus

    qwen/qwen3.7-plus-20260602

    81.8%

    View modelEvidence independent

  7. #7MiniMax M3

    minimax/minimax-m3-20260531

    81.3%

    View modelEvidence independent

  8. #8Qwen3.5-35B-A3B

    qwen/qwen3.5-35b-a3b-20260224

    81.2%

    View modelEvidence independent

  9. #9Qwen3.7 Flash

    qwen/qwen3.7-flash-20260727

    80.8%

    View modelEvidence independent

  10. #10Qwen3.5-9B

    qwen/qwen3.5-9b-20260310

    78.9%

    View modelEvidence independent

  11. #11GPT-5.5

    openai/gpt-5.5-20260423

    77.3%

    View modelEvidence independent

  12. #11GPT-5.6 Terra

    openai/gpt-5.6-terra-20260709

    77.3%

    View modelEvidence independent

  13. #13GPT-5.4

    openai/gpt-5.4-20260305

    74.7%

    View modelEvidence independent

  14. #14DeepSeek V4 Pro 0423

    deepseek/deepseek-v4-pro-20260423

    73.2%

    View modelEvidence independent

  15. #14GPT-5.2

    openai/gpt-5.2-20251211

    73.2%

    View modelEvidence independent

  16. #16GLM 5.2

    z-ai/glm-5.2-20260616

    71.2%

    View modelEvidence independent

  17. #17GPT-5.1

    openai/gpt-5.1-20251113

    66.7%

    View modelEvidence independent

  18. #18GPT-5.4 Mini

    openai/gpt-5.4-mini-20260317

    64.1%

    View modelEvidence independent

  19. #19GPT-5.6 Luna

    openai/gpt-5.6-luna-20260709

    63.6%

    View modelEvidence independent

  20. #20GPT-5.4 Nano

    openai/gpt-5.4-nano-20260317

    55.6%

    View modelEvidence independent

  21. #21Qwen3 32B

    qwen/qwen3-32b-04-28

    54.1%

    View modelEvidence independent

  22. #22Qwen3 14B

    qwen/qwen3-14b-04-28

    53.4%

    View modelEvidence independent

  23. #23Qwen3 30B A3B

    qwen/qwen3-30b-a3b-04-28

    50.4%

    View modelEvidence independent

  24. #24Qwen3 8B

    qwen/qwen3-8b-04-28

    47.5%

    View modelEvidence independent

  25. #25GLM 4.7 Flash

    z-ai/glm-4.7-flash-20260119

    45.1%

    View modelEvidence independent
GPQA Diamond model results | VibeLeaderboard