evals.report
BenchmarksLabsCompareRun guidesIn the wild
AnthropicClaude Opus

Claude Opus 4.6

Anthropic · Claude Opus. Released Feb 5, 2026.

Claude Opus 4.6 is a model from Anthropic in the Claude Opus family, released Feb 5, 2026. evals.report tracks 52 reported Claude Opus 4.6 benchmark scores across FrontierMath, DeepSWE, ARC-AGI-1, ARC-AGI-3, ARC-AGI-2, LMArena, LiveBench, SWE-bench Pro, and 44 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

52 results

Benchmark results 52

Compare this model
BenchmarkCategoryScoreMetricStatusDate
FrontierMathReasoning40.7%accuracyOfficialFeb 5, 2026Details
DeepSWECoding27.06%% resolvedOfficialFeb 5, 2026Details
ARC-AGI-1Reasoning94%accuracyOfficialFeb 5, 2026Details
ARC-AGI-3Reasoning0.51%accuracyOfficialFeb 5, 2026Details
ARC-AGI-2Reasoning69.17%accuracyOfficialFeb 5, 2026Details
LMArenaChat preference1499source-defined ratingOfficialFeb 5, 2026Details
LiveBenchReasoning76.33%scoreOfficialFeb 5, 2026Details
SWE-bench ProCoding51.90%% resolvedOfficialFeb 5, 2026Details
GPQA DiamondReasoning90.5%accuracyOfficialFeb 5, 2026Details
SWE-bench VerifiedCoding78.7%% resolvedOfficialFeb 5, 2026Details
Humanity's Last ExamReasoning34.2%accuracyOfficialFeb 5, 2026Details
MMMU-ProMultimodal77.3%accuracyOfficialFeb 5, 2026Details
AIME (OTIS Mock)Reasoning94.4%accuracyOfficialFeb 5, 2026Details
SimpleQA VerifiedOther46.5%accuracyOfficialFeb 5, 2026Details
GBA EvalCoding44.1%overall scoreOfficialFeb 5, 2026Details
WeirdMLCoding77.9%average accuracyOfficialFeb 5, 2026Details
PostTrainBenchAgents24.82%weighted average scoreOfficialFeb 5, 2026Details
Artificial Analysis Intelligence IndexReasoning53IndexUnverifiedFeb 5, 2026Details
Epoch Capabilities IndexReasoning155.3IndexOfficialFeb 5, 2026Details
SWE-rebenchCoding65.3%Resolved rate (pass@1)OfficialFeb 5, 2026Details
OSWorldAgents72.7%task success rateUnverifiedFeb 5, 2026Details
GAIA: A Benchmark for General AI AssistantsAgents47.8%accuracyUnverifiedFeb 5, 2026Details
τ²-bench (Telecom)Tool use99.3%pass^1UnverifiedFeb 5, 2026Details
AIME 2026Reasoning96.67%accuracyOfficialFeb 5, 2026Details
GDPvalAgents1619EloOfficialFeb 5, 2026Details
METR Task-Completion Time HorizonsAgents718.8 min50% time horizonOfficialFeb 5, 2026Details
SciCodeCoding51.9%accuracyUnverifiedFeb 5, 2026Details
MultiChallengeReasoning56.02%accuracyVerifiedFeb 5, 2026Details
OpenAI-MRCR v2 (Multi-Round Coreference Resolution)Reasoning93.0%accuracy (mean SequenceMatcher similarity)VerifiedFeb 5, 2026Details
Global-MMLUReasoning92.2%accuracyUnverifiedFeb 5, 2026Details
WebDev ArenaChat preference1538EloVerifiedFeb 5, 2026Details
Search ArenaChat preference1251EloVerifiedFeb 5, 2026Details
EQ-Bench Creative Writing v3Chat preference1933EloVerifiedFeb 5, 2026Details
Design ArenaChat preference1344EloVerifiedFeb 5, 2026Details
MASK (Model Alignment between Statements and Knowledge)Other96.28Honesty scoreVerifiedFeb 5, 2026Details
CharXivMultimodal77.4%accuracyUnverifiedFeb 5, 2026Details
OCRBench v2Multimodal48.4accuracyVerifiedFeb 5, 2026Details
ScreenSpot-ProMultimodal83.1%accuracyUnverifiedFeb 5, 2026Details
SWE-bench MultilingualCoding72.0%% resolvedOfficialFeb 5, 2026Details
EnigmaEvalReasoning7.60%accuracyVerifiedFeb 5, 2026Details
ZeroBenchMultimodal11.0% (pass@5)accuracyVerifiedFeb 5, 2026Details
MathArena HMMT February 2026Reasoning96.21%accuracyOfficialFeb 5, 2026Details
FrontierMath Tier 4Reasoning22.9%accuracyOfficialFeb 5, 2026Details
Vectara Hallucination LeaderboardOther12.2%Hallucination RateOfficialFeb 5, 2026Details
Vibe Code BenchCoding57.57%Overall accuracyOfficialFeb 5, 2026Details
WebArenaAgents68.0%Task success rateVerifiedFeb 5, 2026Details
GSO: Software Optimization Benchmark for SWE-AgentsCoding41.18%Opt@1OfficialFeb 5, 2026Details
MultiNRCReasoning57.06%accuracyOfficialFeb 5, 2026Details
Terminal-Bench 2.0Agents65.4%task successVerifiedFeb 5, 2026Details
Remote Labor IndexAgents4.17%automation rateOfficialFeb 5, 2026Details
FrontierSWEAgents56%dominance scoreOfficialFeb 5, 2026Details
ProgramBenchCoding2.5%almost-resolved rateOfficialFeb 5, 2026Details