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 · 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. #1Gemini 3.1 Pro Preview

    google/gemini-3.1-pro-preview-20260219

    94.1%

    View modelEvidence independent

  2. #2Claude Opus 5

    anthropic/claude-opus-5

    92.9%

    View modelEvidence independent

  3. #3MiniMax M3

    minimax/minimax-m3-20260531

    90.9%

    View modelEvidence independent

  4. #3Qwen3.7 Max

    qwen/qwen3.7-max-20260520

    90.9%

    View modelEvidence independent

  5. #5Kimi K2.6

    moonshotai/kimi-k2.6-20260420

    90.8%

    View modelEvidence independent

  6. #6GLM 5.1

    z-ai/glm-5.1-20260406

    89.9%

    View modelEvidence independent

  7. #7Qwen3.6 Plus

    qwen/qwen3.6-plus-04-02

    88.4%

    View modelEvidence independent

  8. #8Kimi K2.7 Code

    moonshotai/kimi-k2.7-code-20260612

    87.9%

    View modelEvidence independent

  9. #8Qwen3.7 Plus

    qwen/qwen3.7-plus-20260602

    87.9%

    View modelEvidence independent

  10. #10GLM 5

    z-ai/glm-5-20260211

    87.8%

    View modelEvidence independent

  11. #11Qwen3.6 Max Preview

    qwen/qwen3.6-max-preview-20260420

    87.4%

    View modelEvidence independent

  12. #12Qwen3.5 397B A17B

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

    85.9%

    View modelEvidence independent

  13. #12Qwen3.6 27B

    qwen/qwen3.6-27b-20260422

    85.9%

    View modelEvidence independent

  14. #14Nemotron 3 Ultra

    nvidia/nemotron-3-ultra-550b-a55b-20260604

    85.4%

    View modelEvidence independent

  15. #15Gemini 2.5 Pro

    google/gemini-2.5-pro

    85.3%

    View modelEvidence independent

  16. #16Gemini 2.5 Pro Preview 06-05

    google/gemini-2.5-pro-preview-06-05

    84.8%

    View modelEvidence independent

  17. #17Qwen3.6 35B A3B

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

    83.8%

    View modelEvidence independent

  18. #18Qwen3.5-35B-A3B

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

    83.5%

    View modelEvidence independent

  19. #19GLM 4.7

    z-ai/glm-4.7-20251222

    83.3%

    View modelEvidence independent

  20. #19Qwen3.6 Flash

    qwen/qwen3.6-flash

    83.3%

    View modelEvidence independent

  21. #21Gemini 3 Flash Preview

    google/gemini-3-flash-preview-20251217

    83.2%

    View modelEvidence independent

  22. #22Qwen3.5-Flash

    qwen/qwen3.5-flash-20260224

    82.3%

    View modelEvidence independent

  23. #22Qwen3.7 Flash

    qwen/qwen3.7-flash-20260727

    82.3%

    View modelEvidence independent

  24. #24Claude Opus 4.5

    anthropic/claude-opus-4-5

    80.7%

    View modelEvidence independent

  25. #25Qwen3 235B A22B Thinking 2507

    qwen/qwen3-235b-a22b-thinking-2507

    80.1%

    View modelEvidence independent

  26. #26Qwen3.5-9B

    qwen/qwen3.5-9b-20260310

    79%

    View modelEvidence independent

  27. #27R1 0528

    deepseek/deepseek-r1-0528

    76.3%

    View modelEvidence independent

  28. #28Claude Sonnet 4.5

    anthropic/claude-sonnet-4-5

    73.7%

    View modelEvidence independent

  29. #29Claude Opus 4.1

    anthropic/claude-opus-4-1

    73.2%

    View modelEvidence independent

  30. #30Qwen3 Max

    qwen/qwen3-max

    72.6%

    View modelEvidence independent

  31. #31R1

    deepseek/deepseek-r1

    71.7%

    View modelEvidence independent

  32. #32Qwen3 235B A22B

    qwen/qwen3-235b-a22b-04-28

    70.7%

    View modelEvidence independent

  33. #33Qwen3 30B A3B Thinking 2507

    qwen/qwen3-30b-a3b-thinking-2507

    70.1%

    View modelEvidence independent

  34. #34Claude Opus 4

    anthropic/claude-opus-4

    69.2%

    View modelEvidence independent

  35. #35DeepSeek V3 0324

    deepseek/deepseek-chat-v3-0324

    67.6%

    View modelEvidence independent

  36. #36GPT-4.1

    openai/gpt-4.1-2025-04-14

    66.9%

    View modelEvidence independent

  37. #37Claude Sonnet 4

    anthropic/claude-sonnet-4

    66.7%

    View modelEvidence independent

  38. #37Gemini 2.5 Pro Preview 05-06

    google/gemini-2.5-pro-preview-03-25

    66.7%

    View modelEvidence independent

  39. #39GPT-4.1 Mini

    openai/gpt-4.1-mini-2025-04-14

    65.8%

    View modelEvidence independent

  40. #40Qwen3 32B

    qwen/qwen3-32b-04-28

    65.7%

    View modelEvidence independent

  41. #41Qwen3 14B

    qwen/qwen3-14b-04-28

    63.8%

    View modelEvidence independent

  42. #42Qwen3 30B A3B

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

    61.7%

    View modelEvidence independent

  43. #43Claude Haiku 4.5

    anthropic/claude-haiku-4-5

    60.5%

    View modelEvidence independent

  44. #43GLM 4.7 Flash

    z-ai/glm-4.7-flash-20260119

    60.5%

    View modelEvidence independent

  45. #45Qwen3 8B

    qwen/qwen3-8b-04-28

    56.8%

    View modelEvidence independent

  46. #46DeepSeek V3

    deepseek/deepseek-chat-v3

    56.5%

    View modelEvidence independent

  47. #47Phi 4

    microsoft/phi-4

    56.1%

    View modelEvidence independent

  48. #48R1 Distill Llama 70B

    deepseek/deepseek-r1-distill-llama-70b

    55.7%

    View modelEvidence independent

  49. #49Qwen3 30B A3B Instruct 2507

    qwen/qwen3-30b-a3b-instruct-2507

    55.6%

    View modelEvidence independent

  50. #50Llama 4 Scout

    meta/llama-4-scout-17b-16e-instruct

    51.8%

    View modelEvidence independent

  51. #51GPT-4o (2024-08-06)

    openai/gpt-4o-2024-08-06

    49.2%

    View modelEvidence independent

  52. #52Qwen2.5 72B Instruct

    qwen/qwen-2.5-72b-instruct

    49.1%

    View modelEvidence independent

  53. #53Mistral Large 2407

    mistralai/mistral-large-2407

    49%

    View modelEvidence independent

  54. #54GPT-4.1 Nano

    openai/gpt-4.1-nano-2025-04-14

    48.9%

    View modelEvidence independent

  55. #54GPT-4o (2024-05-13)

    openai/gpt-4o-2024-05-13

    48.9%

    View modelEvidence independent

  56. #56Qwen-Plus

    qwen/qwen-plus-2025-01-25

    48.1%

    View modelEvidence independent

  57. #57GPT-4o (2024-11-20)

    openai/gpt-4o-2024-11-20

    47.9%

    View modelEvidence independent

  58. #58Gemma 3 27B

    google/gemma-3-27b-it

    47.7%

    View modelEvidence independent

  59. #59Llama 3.3 70B Instruct

    meta/llama-3.3-70b-instruct

    47.4%

    View modelEvidence independent

  60. #60Llama 3.1 70B Instruct

    meta/llama-3.1-70b-instruct

    44.2%

    View modelEvidence independent

  61. #61WizardLM-2 8x22B

    microsoft/wizardlm-2-8x22b

    43.4%

    View modelEvidence independent

  62. #62Gemma 3 12B

    google/gemma-3-12b-it

    39.5%

    View modelEvidence independent

  63. #63GPT-4o-mini (2024-07-18)

    openai/gpt-4o-mini-2024-07-18

    37.7%

    View modelEvidence independent

  64. #64Gemma 2 27B

    google/gemma-2-27b-it

    36.5%

    View modelEvidence independent

  65. #65Claude 3 Haiku

    anthropic/claude-3-haiku

    36.3%

    View modelEvidence independent

  66. #66Qwen2.5 7B Instruct

    qwen/qwen-2.5-7b-instruct

    35.5%

    View modelEvidence independent

  67. #67Granite 4.0 Micro

    ibm-granite/granite-4.0-h-micro

    28.3%

    View modelEvidence independent

  68. #68Llama 3.1 8B Instruct

    meta/llama-3.1-8b-instruct

    27%

    View modelEvidence independent

  69. #69Llama 3.2 1B Instruct

    meta/llama-3.2-1b-instruct

    23.9%

    View modelEvidence independent

  70. #70Gemma 3 4B

    google/gemma-3-4b-it

    23.2%

    View modelEvidence independent
GPQA Diamond model results | VibeLeaderboard