evals.report
BenchmarksLabsCompareRun guidesIn the wild
OpenAIGPT

GPT-5.2

OpenAI · GPT. Released Dec 11, 2025.

GPT-5.2 is a model from OpenAI in the GPT family, released Dec 11, 2025. evals.report tracks 47 reported GPT-5.2 benchmark scores across FrontierMath, Berkeley Function Calling Leaderboard, ARC-AGI-1, LiveBench, LiveCodeBench Pro, SWE-bench Pro, GPQA Diamond, SWE-bench Verified, and 39 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

47 results

Benchmark results 47

Compare this model
BenchmarkCategoryScoreMetricStatusDate
FrontierMathReasoning40.7%accuracyOfficialDec 11, 2025Details
Berkeley Function Calling LeaderboardTool use55.87%accuracyOfficialDec 11, 2025Details
ARC-AGI-1Reasoning86.17%accuracyOfficialDec 11, 2025Details
LiveBenchReasoning74.84%scoreOfficialDec 11, 2025Details
LiveCodeBench ProCoding2393Codeforces EloOfficialDec 11, 2025Details
SWE-bench ProCoding29.94%% resolvedOfficialDec 11, 2025Details
GPQA DiamondReasoning91.4%accuracyOfficialDec 11, 2025Details
SWE-bench VerifiedCoding73.8%% resolvedOfficialDec 11, 2025Details
Humanity's Last ExamReasoning29.9%accuracyOfficialDec 11, 2025Details
MMMU-ProMultimodal80.4%accuracyOfficialDec 11, 2025Details
AIME (OTIS Mock)Reasoning96.1%accuracyOfficialDec 11, 2025Details
SimpleQA VerifiedOther38.9%accuracyOfficialDec 11, 2025Details
ARC-AGI-2Reasoning52.91%accuracyOfficialDec 11, 2025Details
LMArenaChat preference1411source-defined ratingOfficialDec 11, 2025Details
WeirdMLCoding72.2%average accuracyOfficialDec 11, 2025Details
PostTrainBenchAgents21.38%weighted average scoreOfficialDec 11, 2025Details
Artificial Analysis Intelligence IndexReasoning51.3IndexUnverifiedDec 11, 2025Details
Epoch Capabilities IndexReasoning153.7IndexOfficialDec 11, 2025Details
MMLU-ProReasoning85.9%accuracyVerifiedDec 11, 2025Details
OSWorldAgents47.3%task success rateUnverifiedDec 11, 2025Details
GAIA: A Benchmark for General AI AssistantsAgents40.3%accuracyUnverifiedDec 11, 2025Details
τ²-bench (Telecom)Tool use98.7%pass^1UnverifiedDec 11, 2025Details
AIME 2026Reasoning98.33%accuracyOfficialDec 11, 2025Details
GDPvalAgents1467EloOfficialDec 11, 2025Details
LiveCodeBenchCoding89.4%Pass@1UnverifiedDec 11, 2025Details
METR Task-Completion Time HorizonsAgents352.2 min50% time horizonOfficialDec 11, 2025Details
SciCodeCoding46.2%accuracyUnverifiedDec 11, 2025Details
Global-MMLUReasoning89.8%accuracyUnverifiedDec 11, 2025Details
Video-MMMUMultimodal85.9%accuracyUnverifiedDec 11, 2025Details
WebDev ArenaChat preference1404EloVerifiedDec 11, 2025Details
Search ArenaChat preference1210EloVerifiedDec 11, 2025Details
EQ-Bench Creative Writing v3Chat preference1783EloVerifiedDec 11, 2025Details
Design ArenaChat preference1224EloVerifiedDec 11, 2025Details
MASK (Model Alignment between Statements and Knowledge)Other86.67Honesty scoreVerifiedDec 11, 2025Details
CharXivMultimodal82.1%accuracyUnverifiedDec 11, 2025Details
OCRBench v2Multimodal50.5accuracyVerifiedDec 11, 2025Details
ScreenSpot-ProMultimodal86.3%accuracyUnverifiedDec 11, 2025Details
SWE-bench MultilingualCoding66.7%% resolvedOfficialDec 11, 2025Details
EnigmaEvalReasoning10.39%accuracyVerifiedDec 11, 2025Details
ZeroBenchMultimodal17.0% (pass@5)accuracyVerifiedDec 11, 2025Details
MathArena HMMT February 2026Reasoning96.97%accuracyOfficialDec 11, 2025Details
FrontierMath Tier 4Reasoning18.8%accuracyOfficialDec 11, 2025Details
Vectara Hallucination LeaderboardOther10.8%Hallucination RateOfficialDec 11, 2025Details
Vibe Code BenchCoding53.50%Overall accuracyVerifiedDec 11, 2025Details
GSO: Software Optimization Benchmark for SWE-AgentsCoding27.45%Opt@1OfficialDec 11, 2025Details
MultiNRCReasoning42.18%accuracyOfficialDec 11, 2025Details
Remote Labor IndexAgents2.5%automation rateOfficialDec 11, 2025Details