evals.report
BenchmarksLabsCompareRun guidesIn the wild
AnthropicClaude Opus

Claude Opus 4

Anthropic · Claude Opus. Released May 22, 2025.

Claude Opus 4 is a model from Anthropic in the Claude Opus family, released May 22, 2025. evals.report tracks 27 reported Claude Opus 4 benchmark scores across SWE-bench Verified, GPQA Diamond, FrontierMath, AIME (OTIS Mock), LMArena, WeirdML, Epoch Capabilities Index, Aider Polyglot, and 19 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

27 results

Benchmark results 27

Compare this model
BenchmarkCategoryScoreMetricStatusDate
SWE-bench VerifiedCoding70.7%% resolvedOfficialMay 22, 2025Details
GPQA DiamondReasoning76.3%accuracyOfficialMay 22, 2025Details
FrontierMathReasoning4.48%accuracyOfficialMay 22, 2025Details
AIME (OTIS Mock)Reasoning64.4%accuracyOfficialMay 22, 2025Details
LMArenaChat preference1364source-defined ratingOfficialMay 22, 2025Details
WeirdMLCoding43.7%average accuracyOfficialMay 22, 2025Details
Epoch Capabilities IndexReasoning143.4IndexOfficialMay 22, 2025Details
Aider PolyglotCoding72.0%% correctOfficialMay 22, 2025Details
MMLU-ProReasoning87.3%accuracyVerifiedMay 22, 2025Details
GAIA: A Benchmark for General AI AssistantsAgents64.85%accuracyUnverifiedMay 22, 2025Details
LiveCodeBenchCoding54.2%Pass@1UnverifiedMay 22, 2025Details
METR Task-Completion Time HorizonsAgents100.4 min50% time horizonOfficialMay 22, 2025Details
SciCodeCoding40.9%accuracyUnverifiedMay 22, 2025Details
MMMU (Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark)Multimodal76.5%accuracyVerifiedMay 22, 2025Details
MultiChallengeReasoning58.62%accuracyVerifiedMay 22, 2025Details
Search ArenaChat preference1128EloVerifiedMay 22, 2025Details
EQ-Bench Creative Writing v3Chat preference1639EloVerifiedMay 22, 2025Details
Design ArenaChat preference1215EloVerifiedMay 22, 2025Details
MASK (Model Alignment between Statements and Knowledge)Other87.87Honesty scoreVerifiedMay 22, 2025Details
MCP-UniverseTool use28.14%Overall Success RateVerifiedMay 22, 2025Details
EnigmaEvalReasoning5.57%accuracyVerifiedMay 22, 2025Details
ZeroBenchMultimodal1.0% (pass@1)accuracyVerifiedMay 22, 2025Details
IMO-BenchReasoning22.3%accuracyVerifiedMay 22, 2025Details
FrontierMath Tier 4Reasoning4.2%accuracyOfficialMay 22, 2025Details
Vectara Hallucination LeaderboardOther12.0%Hallucination RateOfficialMay 22, 2025Details
GSO: Software Optimization Benchmark for SWE-AgentsCoding6.86%Opt@1OfficialMay 22, 2025Details
MultiNRCReasoning33.93%accuracyOfficialMay 22, 2025Details