evals.report
BenchmarksLabsCompareRun guidesIn the wild
OpenAIGPT

GPT-5

OpenAI · GPT. Released Aug 7, 2025.

GPT-5 is a model from OpenAI in the GPT family, released Aug 7, 2025. evals.report tracks 49 reported GPT-5 benchmark scores across FrontierMath, ARC-AGI-1, LiveCodeBench Pro, SWE-bench Pro, GPQA Diamond, SWE-bench Verified, Humanity's Last Exam, MMMU-Pro, and 41 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

49 results

Benchmark results 49

Compare this model
BenchmarkCategoryScoreMetricStatusDate
FrontierMathReasoning32.41%accuracyOfficialAug 7, 2025Details
ARC-AGI-1Reasoning65.67%accuracyOfficialAug 7, 2025Details
LiveCodeBench ProCoding2176Codeforces EloOfficialAug 7, 2025Details
SWE-bench ProCoding41.78%% resolvedOfficialAug 7, 2025Details
GPQA DiamondReasoning86.2%accuracyOfficialAug 7, 2025Details
SWE-bench VerifiedCoding73.6%% resolvedOfficialAug 7, 2025Details
Humanity's Last ExamReasoning25.32%accuracyOfficialAug 7, 2025Details
MMMU-ProMultimodal78.4%accuracyOfficialAug 7, 2025Details
AIME (OTIS Mock)Reasoning91.4%accuracyOfficialAug 7, 2025Details
SimpleQA VerifiedOther50.6%accuracyOfficialAug 7, 2025Details
ARC-AGI-2Reasoning9.86%accuracyOfficialAug 7, 2025Details
LMArenaChat preference1405source-defined ratingOfficialAug 7, 2025Details
WeirdMLCoding60.7%average accuracyOfficialAug 7, 2025Details
MCP AtlasTool use44.5%pass rateOfficialAug 7, 2025Details
Artificial Analysis Intelligence IndexReasoning44.6IndexUnverifiedAug 7, 2025Details
Epoch Capabilities IndexReasoning150.0IndexOfficialAug 7, 2025Details
Aider PolyglotCoding88.0%% correctOfficialAug 7, 2025Details
MMLU-ProReasoning87.1%accuracyVerifiedAug 7, 2025Details
GAIA: A Benchmark for General AI AssistantsAgents42.1%accuracyUnverifiedAug 7, 2025Details
BrowseCompAgents54.9%accuracyVerifiedAug 7, 2025Details
τ²-bench (Telecom)Tool use96.7%pass^1UnverifiedAug 7, 2025Details
GDPvalAgents1294EloOfficialAug 7, 2025Details
LiveCodeBenchCoding84.6%Pass@1UnverifiedAug 7, 2025Details
METR Task-Completion Time HorizonsAgents203.0 min50% time horizonOfficialAug 7, 2025Details
SciCodeCoding42.9%accuracyUnverifiedAug 7, 2025Details
MMMU (Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark)Multimodal84.2%accuracyVerifiedAug 7, 2025Details
MultiChallengeReasoning63.19%accuracyVerifiedAug 7, 2025Details
Global-MMLUReasoning90.7%accuracyUnverifiedAug 7, 2025Details
Video-MMMUMultimodal84.6%accuracyUnverifiedAug 7, 2025Details
WebDev ArenaChat preference1394EloVerifiedAug 7, 2025Details
Search ArenaChat preference1134EloVerifiedAug 7, 2025Details
EQ-Bench Creative Writing v3Chat preference1640EloVerifiedAug 7, 2025Details
Design ArenaChat preference1223EloVerifiedAug 7, 2025Details
MASK (Model Alignment between Statements and Knowledge)Other79.33Honesty scoreVerifiedAug 7, 2025Details
MCP-UniverseTool use44.16%Overall Success RateVerifiedAug 7, 2025Details
CharXivMultimodal81.1%accuracyUnverifiedAug 7, 2025Details
OCRBench v2Multimodal55.5accuracyVerifiedAug 7, 2025Details
FACTS GroundingReasoning69.6%Grounding accuracyVerifiedAug 7, 2025Details
EnigmaEvalReasoning10.47%accuracyVerifiedAug 7, 2025Details
ZeroBenchMultimodal1.0% (pass@1)accuracyVerifiedAug 7, 2025Details
IMO-BenchReasoning65.6%accuracyVerifiedAug 7, 2025Details
PutnamBenchReasoning28/660Problems solvedVerifiedAug 7, 2025Details
FrontierMath Tier 4Reasoning12.5%accuracyOfficialAug 7, 2025Details
Vectara Hallucination LeaderboardOther15.1%Hallucination RateOfficialAug 7, 2025Details
Gray Swan Arena (Agent Red-Teaming / Indirect Prompt Injection)Agents2.0%Attack Success Rate (ASR)VerifiedAug 7, 2025Details
Vibe Code BenchCoding20.09%Overall accuracyVerifiedAug 7, 2025Details
Online-Mind2WebAgents42.33%Task success rateVerifiedAug 7, 2025Details
GSO: Software Optimization Benchmark for SWE-AgentsCoding6.86%Opt@1OfficialAug 7, 2025Details
MultiNRCReasoning52.13%accuracyOfficialAug 7, 2025Details