supermarket

Vision-language

Claude Sonnet 4.6

By Anthropic, in the Claude Sonnet family.

Released
17 Feb 2026
Status
Legacy
Context
1M
Max output
128K
Input per 1M tokens
$3.00
Output per 1M tokens
$15.00
Open weights
No
License
proprietary
Parameters
—
Knowledge cutoff
Aug 2025
Reasoning
Yes
Modalities
text, image → text

Category scores

Out of 100, from independent and third-party results only, by method category-2026-09-29.1.

Category scores
CategoryScoreRankCoverage
Knowledge & reasoning31.6542 of 4
Coding & software engineering27.1Not ranked2 of 5
Agentic tool use54.9Not ranked1 of 2
Computer & browser use9.361 of 1
Factuality & hallucination35.5421 of 1
Human preference55.2154 of 5

Benchmark results

Every result we hold for this model, with who ran it and where it was published.

Knowledge & reasoning

Knowledge & reasoning results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Chess PuzzlesPuzzles solved13.0%Independent32K20 Feb 2026Epoch AI
Chess PuzzlesPuzzles solved8.0%Independentmedium13 Jul 2026Epoch AI
Chess PuzzlesPuzzles solved5.0%Independenthigh13 Jul 2026Epoch AI
Chess PuzzlesPuzzles solved3.0%Independentmax6 Aug 2026Epoch AI
DTBench (Decision Theory Benchmark)Accuracy (capability questions)89.9%Third partymax—https://conceptualreasoning.ai/dtbench via Epoch AI
LMCA (Language Model Conceptual Argumentation)Correlation with expert ratings (Pearson r)0.4651Third partymax—https://conceptualreasoning.ai/lmca via Epoch AI
Mystery Game PuzzlesPuzzles solved16.0%Independentlow6 Aug 2026Epoch AI
Mystery Game PuzzlesPuzzles solved14.0%Independent—6 Aug 2026Epoch AI

Scientific reasoning

Scientific reasoning results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
GPQA DiamondAccuracy87.4%Independent32K20 Feb 2026Epoch AI
GPQA DiamondAccuracy83.3%Independentmedium13 Jul 2026Epoch AI
GPQA DiamondAccuracy83.3%Independenthigh13 Jul 2026Epoch AI
GPQA DiamondAccuracy78.8%Independentmax6 Aug 2026Epoch AI

Mathematics

Mathematics results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
FrontierMath Tier 4 (v1)Accuracy8.3%Independent16K21 Feb 2026Epoch AI
FrontierMath Tiers 1-3 (v1)Accuracy32.4%Independent16K20 Feb 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy85.8%Independent32K20 Feb 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy82.2%Independentmedium13 Jul 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy75.6%Independenthigh13 Jul 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy71.1%Independentmax6 Aug 2026Epoch AI

Coding & software engineering

Coding & software engineering results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
DeepSWEPass@129.9%Third partyhigh—https://deepswe.datacurve.ai/ via Epoch AI
ExploitBenchMean capability23.6%Source not stated——Epoch AI via Epoch AI
ExploitBenchMean capability22.7%Source not stated——Epoch AI via Epoch AI
FrontierCodeMain score (Mean@5)24.3%Third partyunknown—https://cognition.com/frontiercode via Epoch AI
SWE-bench Verified% resolved75.2%Independent—21 Feb 2026Epoch AI
Terminal-Bench 2.0Accuracy (mean over trials)53.4%Third partyunknown—https://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI
Terminal-Bench 2.0Accuracy (mean over trials)53.4%Source not stated——Epoch AI via Epoch AI
WeirdML (v2)Average accuracy66.1%Third partymedium—WeirdML Leaderboard via Epoch AI

Agentic tool use

Agentic tool use results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Deep Research Bench (FutureSearch)Average score54.9%Third partyhigh—https://evals.futuresearch.ai/#drb via Epoch AI
Deep Research Bench (FutureSearch)Average score50.4%Third partylow—https://evals.futuresearch.ai/#drb via Epoch AI

Computer & browser use

Computer & browser use results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
OSWorld 2.0Binary completion (500 steps)9.3%Third partymedium—https://osworld-v2.xlang.ai/ via Epoch AI
OSWorld 2.0Binary completion (500 steps)8.3%Third partymax—https://osworld-v2.xlang.ai/ via Epoch AI
OSWorld-Verified% tasks completed72.1%Third party——OS World Website via Epoch AI

Professional knowledge work

Professional knowledge work results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
APEX-AgentsPass@1 (all rubric criteria met)43.0%Source not statedhigh—Epoch AI via Epoch AI

Factuality & hallucination

Factuality & hallucination results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
SimpleQA VerifiedAccuracy (% correct)35.5%Independenthigh10 Aug 2026Epoch AI
SimpleQA VerifiedAccuracy (% correct)32.8%Independentmax27 Aug 2026Epoch AI

Human preference

Human preference results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Code Arena: WebDev (arena.ai)Arena score (Elo scale)1521Independent—25 Sep 2026Arena
Code Arena: WebDev (arena.ai)Votes22980Independent—25 Sep 2026Arena
Document Arena (arena.ai)Arena score (Elo scale)1478Independent—13 Sep 2026Arena
Document Arena (arena.ai)Votes57813Independent—13 Sep 2026Arena
Text Arena (arena.ai)Arena score (Elo scale)1472Independent—25 Sep 2026Arena
Text Arena (arena.ai)Votes70662Independent—25 Sep 2026Arena
Vision Arena (arena.ai)Arena score (Elo scale)1275Independent—28 Sep 2026Arena
Vision Arena (arena.ai)Votes27465Independent—28 Sep 2026Arena

Other Claude Sonnet versions

Sources

Claude Sonnet 4.6: benchmarks, price and specs · supermarket.dev