Acropolis

GPT-5.5 Pro Pre-release

OpenAIGrade C. Evidence grade C of A to D: thin.ProprietaryReleased 23 Apr 2026
Compare with peers

Benchmark appearances

Best verified run per board, with confidence intervals and Pareto frontier status.

BenchmarkRankScoreCostStatusObserved
FrontierMath-2025-02-28-Private

Epoch AI Benchmarking Hub · high

#1 / 101Tied #1
52.4%[46.7% to 58.1%]UnknownStandard23 Apr 2026
FrontierMath-2025-02-28-Private

Epoch AI Benchmarking Hub · xhigh

#3 / 101Tied #1
51%[45.3% to 56.7%]UnknownStandard23 Apr 2026
GPQA diamond

Epoch AI Benchmarking Hub · xhigh

#19 / 338Tied #1
93.9%[90.8% to 97%]UnknownStandard24 Apr 2026

Reasoning effort scaling

Same benchmark and harness measured across multiple reasoning effort settings.

  • FrontierMath-2025-02-28-Private

    Epoch AI Benchmarking Hub

    1 regression. Evidence grade D of A to D: weakest.
    high

    52.4

    xhigh

    51.0