evals.report
BenchmarksLabsCompareRun guidesIn the wild
AnthropicClaude Sonnet

Claude Sonnet 4.6

Anthropic · Claude Sonnet. Released Feb 17, 2026.

Claude Sonnet 4.6 is a model from Anthropic in the Claude Sonnet family, released Feb 17, 2026. evals.report tracks 38 reported Claude Sonnet 4.6 benchmark scores across FrontierMath, DeepSWE, ARC-AGI-1, LMArena, LiveBench, GPQA Diamond, SWE-bench Verified, Humanity's Last Exam, and 30 more — each shown with its benchmark, metric, source status, and date, and never combined into a single ranking.

38 results

Benchmark results 38

Compare this model
BenchmarkCategoryScoreMetricStatusDate
FrontierMathReasoning32.4%accuracyOfficialFeb 17, 2026Details
DeepSWECoding31.56%% resolvedOfficialFeb 17, 2026Details
ARC-AGI-1Reasoning86.5%accuracyOfficialFeb 17, 2026Details
LMArenaChat preference1454source-defined ratingOfficialFeb 17, 2026Details
LiveBenchReasoning75.47%scoreOfficialFeb 17, 2026Details
GPQA DiamondReasoning87.4%accuracyOfficialFeb 17, 2026Details
SWE-bench VerifiedCoding75.2%% resolvedOfficialFeb 17, 2026Details
Humanity's Last ExamReasoning21.07%accuracyOfficialFeb 17, 2026Details
MMMU-ProMultimodal75.6%accuracyOfficialFeb 17, 2026Details
AIME (OTIS Mock)Reasoning85.8%accuracyOfficialFeb 17, 2026Details
SimpleQA VerifiedOther29.0%accuracyOfficialFeb 17, 2026Details
ARC-AGI-2Reasoning60.42%accuracyOfficialFeb 17, 2026Details
GBA EvalCoding48.8%overall scoreOfficialFeb 17, 2026Details
WeirdMLCoding66.1%average accuracyOfficialFeb 17, 2026Details
CursorBenchAgents49.0%scoreOfficialFeb 17, 2026Details
PostTrainBenchAgents16.42%weighted average scoreOfficialFeb 17, 2026Details
Artificial Analysis Intelligence IndexReasoning44.4IndexUnverifiedFeb 17, 2026Details
Epoch Capabilities IndexReasoning152.6IndexOfficialFeb 17, 2026Details
SWE-rebenchCoding60.7%Resolved rate (pass@1)UnverifiedFeb 17, 2026Details
OSWorldAgents72.1%task success rateUnverifiedFeb 17, 2026Details
GAIA: A Benchmark for General AI AssistantsAgents45.5%accuracyUnverifiedFeb 17, 2026Details
τ²-bench (Telecom)Tool use75.7%pass^1OfficialFeb 17, 2026Details
GDPvalAgents1676EloOfficialFeb 17, 2026Details
SciCodeCoding46.9%accuracyUnverifiedFeb 17, 2026Details
AA-Omniscience: Knowledge and Hallucination BenchmarkReasoning12AA-Omniscience IndexOfficialFeb 17, 2026Details
IFBenchReasoning56.6%accuracyOfficialFeb 17, 2026Details
Global-MMLUReasoning90.5%accuracyUnverifiedFeb 17, 2026Details
WebDev ArenaChat preference1523EloVerifiedFeb 17, 2026Details
Search ArenaChat preference1219EloVerifiedFeb 17, 2026Details
EQ-Bench Creative Writing v3Chat preference1968EloVerifiedFeb 17, 2026Details
Design ArenaChat preference1327EloVerifiedFeb 17, 2026Details
FrontierMath Tier 4Reasoning8.3%accuracyOfficialFeb 17, 2026Details
Vectara Hallucination LeaderboardOther10.6%Hallucination RateOfficialFeb 17, 2026Details
Vibe Code BenchCoding51.48%Overall accuracyVerifiedFeb 17, 2026Details
WebArenaAgents65.6%Task success rateVerifiedFeb 17, 2026Details
Terminal-Bench 2.0Agents59.1%task successVerifiedFeb 17, 2026Details
ProgramBenchCoding1.0%almost-resolved rateOfficialFeb 17, 2026Details
FrontierCodeCoding3.5%weighted score (Diamond)OfficialFeb 17, 2026Details