evals.report
BenchmarksLabsCompareRun guidesIn the wild
OpenAIGPT

GPT-5.5

OpenAI · GPT. Released Apr 23, 2026.

GPT-5.5 is a model from OpenAI in the GPT family, released Apr 23, 2026. evals.report tracks 41 reported GPT-5.5 benchmark scores across FrontierMath, DeepSWE, ARC-AGI-1, ARC-AGI-3, ARC-AGI-2, LMArena, LiveBench, SWE-bench Pro, and 33 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

41 results

Benchmark results 41

Compare this model
BenchmarkCategoryScoreMetricStatusDate
FrontierMathReasoning51.7%accuracyOfficialApr 23, 2026Details
DeepSWECoding70.05%% resolvedOfficialApr 23, 2026Details
ARC-AGI-1Reasoning95%accuracyOfficialApr 23, 2026Details
ARC-AGI-3Reasoning0.43%accuracyOfficialApr 23, 2026Details
ARC-AGI-2Reasoning85%accuracyOfficialApr 23, 2026Details
LMArenaChat preference1468source-defined ratingOfficialApr 23, 2026Details
LiveBenchReasoning80.71%scoreOfficialApr 23, 2026Details
SWE-bench ProCoding58.6%% resolvedVerifiedApr 23, 2026Details
GPQA DiamondReasoning94.0%accuracyOfficialApr 23, 2026Details
SWE-bench VerifiedCoding80.6%% resolvedOfficialApr 23, 2026Details
Humanity's Last ExamReasoning43.56%accuracyOfficialApr 23, 2026Details
AIME (OTIS Mock)Reasoning100.0%accuracyOfficialApr 23, 2026Details
SimpleQA VerifiedOther63.1%accuracyOfficialApr 23, 2026Details
GBA EvalCoding53.2%overall scoreOfficialApr 23, 2026Details
WeirdMLCoding84.9%average accuracyOfficialApr 23, 2026Details
CursorBenchAgents64.3%scoreOfficialApr 23, 2026Details
PostTrainBenchAgents25.02%weighted average scoreOfficialApr 23, 2026Details
Artificial Analysis Intelligence IndexReasoning60IndexOfficialApr 23, 2026Details
Epoch Capabilities IndexReasoning158.2IndexOfficialApr 23, 2026Details
OSWorldAgents78.7%task success rateUnverifiedApr 23, 2026Details
τ²-bench (Telecom)Tool use93.9%pass^1OfficialApr 23, 2026Details
AIME 2026Reasoning97.50%accuracyOfficialApr 23, 2026Details
GDPvalAgents1769EloOfficialApr 23, 2026Details
AA-Omniscience: Knowledge and Hallucination BenchmarkReasoning20AA-Omniscience IndexOfficialApr 23, 2026Details
IFBenchReasoning75.9%accuracyOfficialApr 23, 2026Details
OpenAI-MRCR v2 (Multi-Round Coreference Resolution)Reasoning74.0%accuracy (mean SequenceMatcher similarity)VerifiedApr 23, 2026Details
WebDev ArenaChat preference1505EloVerifiedApr 23, 2026Details
Search ArenaChat preference1239EloVerifiedApr 23, 2026Details
EQ-Bench Creative Writing v3Chat preference2035EloVerifiedApr 23, 2026Details
Design ArenaChat preference1301EloVerifiedApr 23, 2026Details
CharXivMultimodal84.1%accuracyUnverifiedApr 23, 2026Details
ZeroBenchMultimodal22.0% (pass@5)accuracyVerifiedApr 23, 2026Details
MathArena HMMT February 2026Reasoning97.73%accuracyOfficialApr 23, 2026Details
FrontierMath Tier 4Reasoning35.4%accuracyOfficialApr 23, 2026Details
Vectara Hallucination LeaderboardOther9.3%Hallucination RateOfficialApr 23, 2026Details
Vibe Code BenchCoding69.85%Overall accuracyOfficialApr 23, 2026Details
GSO: Software Optimization Benchmark for SWE-AgentsCoding40.2%Opt@1OfficialApr 23, 2026Details
Remote Labor IndexAgents6.25%automation rateOfficialApr 23, 2026Details
FrontierSWEAgents73%dominance scoreOfficialApr 23, 2026Details
ProgramBenchCoding13.5%almost-resolved rateOfficialApr 23, 2026Details
FrontierCodeCoding6.3%weighted score (Diamond)OfficialApr 23, 2026Details