Acropolis

GPT-5.5 Pre-release

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

Domain profile

Median within-board percentile by capability domain across boards with sufficient peer overlap.

  • Agentic (1 boards): 97 (scale 0 to 100)Agentic (1 boards)97
  • Coding (1 boards): 97 (scale 0 to 100)Coding (1 boards)97
  • Math (1 boards): 99 (scale 0 to 100)Math (1 boards)99
  • Science (1 boards): 93 (scale 0 to 100)Science (1 boards)93

Benchmark appearances

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

BenchmarkRankScoreCostStatusObserved
FrontierMath-2025-02-28-Private

Epoch AI Benchmarking Hub · xhigh

#2 / 101Tied #1
51.7%[46% to 57.4%]UnknownStandard23 Apr 2026
SWE-Bench verified

Epoch AI Benchmarking Hub · xhigh

#2 / 35Tied #1
80.6%[77.1% to 84.1%]UnknownStandard24 Apr 2026
GPQA diamond

Epoch AI Benchmarking Hub · xhigh

#17 / 338Tied #1
94%[91% to 97%]UnknownStandard24 Apr 2026

Closest head-to-head rivals

Direct pairwise win rates for GPT-5.5 Pre-release across shared benchmark boards.