evals.report
BenchmarksLabsCompareRun guidesIn the wild
AnthropicClaude Opus

Claude Opus 4.7

Anthropic · Claude Opus. Released Apr 16, 2026.

Claude Opus 4.7 is a model from Anthropic in the Claude Opus family, released Apr 16, 2026. evals.report tracks 40 reported Claude Opus 4.7 benchmark scores across FrontierMath, DeepSWE, ARC-AGI-1, ARC-AGI-3, ARC-AGI-2, LMArena, LiveBench, GPQA Diamond, and 32 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

40 results

Benchmark results 40

Compare this model
BenchmarkCategoryScoreMetricStatusDate
FrontierMathReasoning43.79%accuracyOfficialApr 16, 2026Details
DeepSWECoding54.20%% resolvedOfficialApr 16, 2026Details
ARC-AGI-1Reasoning92%accuracyOfficialApr 16, 2026Details
ARC-AGI-3Reasoning0.18%accuracyOfficialApr 16, 2026Details
ARC-AGI-2Reasoning75.83%accuracyOfficialApr 16, 2026Details
LMArenaChat preference1486source-defined ratingOfficialApr 16, 2026Details
LiveBenchReasoning76.91%scoreOfficialApr 16, 2026Details
GPQA DiamondReasoning90.2%accuracyOfficialApr 16, 2026Details
SWE-bench VerifiedCoding83.5%% resolvedOfficialApr 16, 2026Details
Humanity's Last ExamReasoning39.04%accuracyOfficialApr 16, 2026Details
AIME (OTIS Mock)Reasoning97.8%accuracyOfficialApr 16, 2026Details
SimpleQA VerifiedOther50.6%accuracyOfficialApr 16, 2026Details
GBA EvalCoding43.8%overall scoreOfficialApr 16, 2026Details
WeirdMLCoding76.4%average accuracyOfficialApr 16, 2026Details
CursorBenchAgents64.8%scoreOfficialApr 16, 2026Details
PostTrainBenchAgents28.56%weighted average scoreOfficialApr 16, 2026Details
Artificial Analysis Intelligence IndexReasoning57IndexOfficialApr 16, 2026Details
Epoch Capabilities IndexReasoning156.2IndexOfficialApr 16, 2026Details
MMLU-ProReasoning89.87%accuracyVerifiedApr 16, 2026Details
OSWorldAgents82.8%task success rateVerifiedApr 16, 2026Details
τ²-bench (Telecom)Tool use88.6%pass^1OfficialApr 16, 2026Details
AIME 2026Reasoning95.83%accuracyOfficialApr 16, 2026Details
GDPvalAgents1753EloOfficialApr 16, 2026Details
SciCodeCoding54.5%accuracyUnverifiedApr 16, 2026Details
AA-Omniscience: Knowledge and Hallucination BenchmarkReasoning26AA-Omniscience IndexOfficialApr 16, 2026Details
IFBenchReasoning58.6%accuracyOfficialApr 16, 2026Details
WebDev ArenaChat preference1562EloVerifiedApr 16, 2026Details
Search ArenaChat preference1237EloVerifiedApr 16, 2026Details
EQ-Bench Creative Writing v3Chat preference2206EloVerifiedApr 16, 2026Details
Design ArenaChat preference1328EloVerifiedApr 16, 2026Details
CharXivMultimodal91.0%accuracyUnverifiedApr 16, 2026Details
ZeroBenchMultimodal14.0% (pass@5)accuracyVerifiedApr 16, 2026Details
MathArena HMMT February 2026Reasoning93.94%accuracyOfficialApr 16, 2026Details
FrontierMath Tier 4Reasoning22.9%accuracyOfficialApr 16, 2026Details
Vectara Hallucination LeaderboardOther12.0%Hallucination RateOfficialApr 16, 2026Details
Vibe Code BenchCoding71.00%Overall accuracyOfficialApr 16, 2026Details
GSO: Software Optimization Benchmark for SWE-AgentsCoding44.12%Opt@1OfficialApr 16, 2026Details
FrontierSWEAgents63%dominance scoreOfficialApr 16, 2026Details
ProgramBenchCoding4.5%almost-resolved rateOfficialApr 16, 2026Details
FrontierCodeCoding5.2%weighted score (Diamond)OfficialApr 16, 2026Details