Acropolis

HLE

Humanity's Last Exam results that Epoch AI republishes from the original runner.

Upstream source
ScienceMathGeneral text
Independent. Run by an operator independent of the model vendors.Epoch AI Benchmarking Hub by Epoch AIVersion 1Data as of 17 Nov 2026

No filters. Showing 56 of 56 rows.

  1. 1
    OpenAIDossier

    Accuracy

    GPT-6 Astra, Accuracy: 54.8% (scale 0.0 to 100.0), 95% confidence interval 51.0 to 58.654.8%(51.0–58.6)
    Calibration error
    39%
  2. 2
    AnthropicDossier

    xhigh

    Accuracy

    Fable 5.1, Accuracy: 46.5% (scale 0.0 to 100.0), 95% confidence interval 42.6 to 50.446.5%(42.6–50.4)
    Calibration error
    20%
  3. 3
    GoogleDossier

    high

    Accuracy

    Gemini 3.1 Pro Preview, Accuracy: 46.4% (scale 0.0 to 100.0), 95% confidence interval 42.6 to 50.346.4%(42.6–50.3)
    Calibration error
    51%
  4. 4
    AnthropicDossier

    xhigh

    Accuracy

    Sonnet 5.5, Accuracy: 46.1% (scale 0.0 to 100.0), 95% confidence interval 42.3 to 49.946.1%(42.3–49.9)
    Calibration error
    Not tested
  5. 5
    GoogleDossier

    Accuracy

    Gemini 3.8 Flash, Accuracy: 44.5% (scale 0.0 to 100.0), 95% confidence interval 40.7 to 48.444.5%(40.7–48.4)
    Calibration error
    51%
  6. 6
    OpenAIDossier

    Accuracy

    GPT-5.4 Pro, Accuracy: 44.3% (scale 0.0 to 100.0), 95% confidence interval 40.5 to 48.144.3%(40.5–48.1)
    Calibration error
    38%
  7. 7

    Accuracy

    Muse Spark, Accuracy: 40.6% (scale 0.0 to 100.0), 95% confidence interval 36.8 to 44.340.6%(36.8–44.3)
    Calibration error
    50%
  8. 8
    GoogleDossier

    Accuracy

    Gemini 3 Pro Preview, Accuracy: 37.5% (scale 0.0 to 100.0), 95% confidence interval 33.8 to 41.237.5%(33.8–41.2)
    Calibration error
    57%
  9. 9
    OpenAIDossier

    xhigh

    Accuracy

    GPT-5.4, Accuracy: 36.2% (scale 0.0 to 100.0), 95% confidence interval 32.6 to 39.936.2%(32.6–39.9)
    Calibration error
    42%
  10. 10
    AnthropicDossier

    Accuracy

    Opus 4.7, Accuracy: 36.2% (scale 0.0 to 100.0), 95% confidence interval 32.5 to 39.936.2%(32.5–39.9)
    Calibration error
    47%
  11. 11
    AnthropicDossier

    max

    Accuracy

    Opus 4.6, Accuracy: 34.4% (scale 0.0 to 100.0), 95% confidence interval 30.8 to 38.134.4%(30.8–38.1)
    Calibration error
    46%
  12. 12
    OpenAIDossier

    Accuracy

    GPT-5 Pro, Accuracy: 31.6% (scale 0.0 to 100.0), 95% confidence interval 28.1 to 35.231.6%(28.1–35.2)
    Calibration error
    49%
  13. 13
    AnthropicDossier

    Accuracy

    Haiku 5.5, Accuracy: 30.1% (scale 0.0 to 100.0), 95% confidence interval 26.5 to 33.730.1%(26.5–33.7)
    Calibration error
    Not tested
  14. 14
    OpenAIDossier

    Accuracy

    GPT-5.2, Accuracy: 27.8% (scale 0.0 to 100.0), 95% confidence interval 24.4 to 31.227.8%(24.4–31.2)
    Calibration error
    45%
  15. 15
    OpenAIDossier

    high

    Accuracy

    GPT-5, Accuracy: 25.3% (scale 0.0 to 100.0), 95% confidence interval 22.0 to 28.725.3%(22.0–28.7)
    Calibration error
    50%
  16. 15
    OpenAIDossier

    Accuracy

    GPT-5, Accuracy: 25.3% (scale 0.0 to 100.0), 95% confidence interval 22.0 to 28.725.3%(22.0–28.7)
    Calibration error
    50%
  17. 17
    AnthropicDossier

    Accuracy

    Opus 4.5, Accuracy: 25.2% (scale 0.0 to 100.0), 95% confidence interval 21.9 to 28.525.2%(21.9–28.5)
    Calibration error
    55%
  18. 18
    Moonshot AIDossier

    Accuracy

    Kimi K2.5, Accuracy: 24.4% (scale 0.0 to 100.0), 95% confidence interval 20.8 to 27.924.4%(20.8–27.9)
    Calibration error
    67%
  19. 19
    OpenAIDossier

    Accuracy

    GPT-5.1, Accuracy: 23.7% (scale 0.0 to 100.0), 95% confidence interval 20.4 to 27.023.7%(20.4–27.0)
    Calibration error
    55%
  20. 20
    GoogleDossier

    Accuracy

    Gemini 2.5 Pro Preview (06-05), Accuracy: 21.6% (scale 0.0 to 100.0), 95% confidence interval 18.5 to 24.821.6%(18.5–24.8)
    Calibration error
    72%
  21. 21
    OpenAIDossier

    high

    Accuracy

    o3, Accuracy: 20.3% (scale 0.0 to 100.0), 95% confidence interval 17.2 to 23.420.3%(17.2–23.4)
    Calibration error
    34%
  22. 22
    OpenAIDossier

    Accuracy

    GPT-5 mini, Accuracy: 19.4% (scale 0.0 to 100.0), 95% confidence interval 16.4 to 22.519.4%(16.4–22.5)
    Calibration error
    65%
  23. 23
    OpenAIDossier

    medium

    Accuracy

    o3, Accuracy: 19.2% (scale 0.0 to 100.0), 95% confidence interval 16.2 to 22.219.2%(16.2–22.2)
    Calibration error
    39%
  24. 24
    AnthropicDossier

    Accuracy

    Opus 4.6, Accuracy: 19% (scale 0.0 to 100.0), 95% confidence interval 16.0 to 22.019%(16.0–22.0)
    Calibration error
    44%
  25. 25
    GoogleDossier

    Accuracy

    Gemini 2.5 Pro Experimental (03-25), Accuracy: 18.2% (scale 0.0 to 100.0), 95% confidence interval 15.2 to 21.118.2%(15.2–21.1)
    Calibration error
    71%
  26. 26
    OpenAIDossier

    high

    Accuracy

    o4-mini, Accuracy: 18.1% (scale 0.0 to 100.0), 95% confidence interval 15.1 to 21.018.1%(15.1–21.0)
    Calibration error
    57%
  27. 27
    GoogleDossier

    Accuracy

    Gemini 2.5 Pro Preview (May '25), Accuracy: 17.8% (scale 0.0 to 100.0), 95% confidence interval 14.9 to 20.717.8%(14.9–20.7)
    Calibration error
    70%
  28. 28
    OpenAIDossier

    medium

    Accuracy

    o4-mini, Accuracy: 14.3% (scale 0.0 to 100.0), 95% confidence interval 11.6 to 17.014.3%(11.6–17.0)
    Calibration error
    59%
  29. 29
    AnthropicDossier

    Accuracy

    Opus 4.5, Accuracy: 14.2% (scale 0.0 to 100.0), 95% confidence interval 11.5 to 16.814.2%(11.5–16.8)
    Calibration error
    56%
  30. 30
    AnthropicDossier

    Accuracy

    Sonnet 4.5, Accuracy: 13.7% (scale 0.0 to 100.0), 95% confidence interval 11.1 to 16.413.7%(11.1–16.4)
    Calibration error
    65%
  31. 31
    GoogleDossier

    Accuracy

    Gemini 2.5 Flash Preview (04-17), Accuracy: 12.1% (scale 0.0 to 100.0), 95% confidence interval 9.6 to 14.612.1%(9.6–14.6)
    Calibration error
    80%
  32. 32
    AnthropicDossier

    Accuracy

    Opus 4.1, Accuracy: 11.5% (scale 0.0 to 100.0), 95% confidence interval 9.1 to 14.011.5%(9.1–14.0)
    Calibration error
    71%
  33. 33
    GoogleDossier

    Accuracy

    Gemini 2.5 Flash Preview (05-20), Accuracy: 11% (scale 0.0 to 100.0), 95% confidence interval 8.6 to 13.411%(8.6–13.4)
    Calibration error
    82%
  34. 34
    AnthropicDossier

    thinking

    Accuracy

    Opus 4, Accuracy: 10.7% (scale 0.0 to 100.0), 95% confidence interval 8.3 to 13.110.7%(8.3–13.1)
    Calibration error
    73%
  35. 35
    GoogleDossier

    Accuracy

    Gemini 3.1 Flash-Lite Preview, Accuracy: 8.6% (scale 0.0 to 100.0), 95% confidence interval 6.5 to 10.88.6%(6.5–10.8)
    Calibration error
    83%
  36. 36

    Accuracy

    GLM-4.5, Accuracy: 8.3% (scale 0.0 to 100.0), 95% confidence interval 6.2 to 10.48.3%(6.2–10.4)
    Calibration error
    79%
  37. 37

    Accuracy

    GLM-4.5-Air, Accuracy: 8.1% (scale 0.0 to 100.0), 95% confidence interval 6.0 to 10.28.1%(6.0–10.2)
    Calibration error
    77%
  38. 37
    OpenAIDossier

    Accuracy

    o1-pro, Accuracy: 8.1% (scale 0.0 to 100.0), 95% confidence interval 6.0 to 10.28.1%(6.0–10.2)
    Calibration error
    82%
  39. 39
    AnthropicDossier

    thinking

    Accuracy

    Claude 3.7 Sonnet, Accuracy: 8% (scale 0.0 to 100.0), 95% confidence interval 5.9 to 10.18%(5.9–10.1)
    Calibration error
    80%
  40. 40
    OpenAIDossier

    Accuracy

    o1, Accuracy: 8% (scale 0.0 to 100.0), 95% confidence interval 5.9 to 10.08%(5.9–10.0)
    Calibration error
    83%
  41. 41
    AnthropicDossier

    Accuracy

    Opus 4.1, Accuracy: 7.9% (scale 0.0 to 100.0), 95% confidence interval 5.8 to 10.07.9%(5.8–10.0)
    Calibration error
    70%
  42. 42
    AnthropicDossier

    thinking

    Accuracy

    Claude Sonnet 4, Accuracy: 7.8% (scale 0.0 to 100.0), 95% confidence interval 5.7 to 9.87.8%(5.7–9.8)
    Calibration error
    75%
  43. 43
    AnthropicDossier

    Accuracy

    Sonnet 4.5, Accuracy: 7.5% (scale 0.0 to 100.0), 95% confidence interval 5.5 to 9.57.5%(5.5–9.5)
    Calibration error
    70%
  44. 44
    OpenAIDossier

    none

    Accuracy

    GPT-5.1, Accuracy: 6.8% (scale 0.0 to 100.0), 95% confidence interval 4.9 to 8.76.8%(4.9–8.7)
    Calibration error
    69%
  45. 45
    AnthropicDossier

    Accuracy

    Opus 4, Accuracy: 6.7% (scale 0.0 to 100.0), 95% confidence interval 4.8 to 8.66.7%(4.8–8.6)
    Calibration error
    74%
  46. 46
    GoogleDossier

    Accuracy

    Gemini 2.0 Flash Thinking Experimental (01-21), Accuracy: 6.6% (scale 0.0 to 100.0), 95% confidence interval 4.7 to 8.56.6%(4.7–8.5)
    Calibration error
    82%
  47. 47

    Accuracy

    Llama 4 Maverick, Accuracy: 5.7% (scale 0.0 to 100.0), 95% confidence interval 3.9 to 7.55.7%(3.9–7.5)
    Calibration error
    83%
  48. 48
    AnthropicDossier

    Accuracy

    Claude Sonnet 4, Accuracy: 5.5% (scale 0.0 to 100.0), 95% confidence interval 3.8 to 7.35.5%(3.8–7.3)
    Calibration error
    76%
  49. 49
    OpenAIDossier

    Accuracy

    GPT-4.5 Preview, Accuracy: 5.4% (scale 0.0 to 100.0), 95% confidence interval 3.7 to 7.25.4%(3.7–7.2)
    Calibration error
    85%
  50. 50
    OpenAIDossier

    Accuracy

    GPT-4.1, Accuracy: 5.4% (scale 0.0 to 100.0), 95% confidence interval 3.7 to 7.15.4%(3.7–7.1)
    Calibration error
    89%
  51. 51
    GoogleDossier

    Accuracy

    Gemini 1.5 Pro (002), Accuracy: 4.6% (scale 0.0 to 100.0), 95% confidence interval 3.0 to 6.24.6%(3.0–6.2)
    Calibration error
    88%
  52. 52
    MistralDossier

    Accuracy

    Mistral Medium 3, Accuracy: 4.5% (scale 0.0 to 100.0), 95% confidence interval 2.9 to 6.14.5%(2.9–6.1)
    Calibration error
    77%
  53. 53
    AmazonDossier

    Accuracy

    Nova Pro, Accuracy: 4.4% (scale 0.0 to 100.0), 95% confidence interval 2.8 to 6.04.4%(2.8–6.0)
    Calibration error
    80%
  54. 54
    AnthropicDossier

    Accuracy

    Claude 3.5 Sonnet (Oct '24), Accuracy: 4.1% (scale 0.0 to 100.0), 95% confidence interval 2.6 to 5.64.1%(2.6–5.6)
    Calibration error
    84%
  55. 55
    AmazonDossier

    Accuracy

    Nova Lite, Accuracy: 3.6% (scale 0.0 to 100.0), 95% confidence interval 2.2 to 5.13.6%(2.2–5.1)
    Calibration error
    82%
  56. 56
    OpenAIDossier

    Accuracy

    GPT-4o (Nov '24), Accuracy: 2.7% (scale 0.0 to 100.0), 95% confidence interval 1.5 to 4.02.7%(1.5–4.0)
    Calibration error
    89%