Model

Claude Sonnet 4.6

Release Date: Feb 17, 2026

Developer Anthropic 🇺🇸

Context Window 1M

Max Output Tokens 128k

Token Costs (in/out) $3.00/15.00

Weights Private

Input Modalities

Accuracy

58.54 %

Avg. Cost (In/Out)

$ 3.00 / $ 15.00

Latency

18 min 37 s

BenchmarksAccuracyRankings

Code Migration

39.89%

±4.14

27/73

EMB

60.15%

±3.28

31/70

Finance Agent (v2)

51.03%

±0.33

35/74

Legal Research Bench

38.46%

±3.38

30/73

MortgageTax

67.73%

±0.92

22/98

SAGE

46.58%

±3.43

34/90

Public Benefits Bench v1.1

62.45%

±1.26

24/47

Tax Agent Bench

64.85%

±3.11

20/65

TaxEval v2

77.11%

±0.82

4/145

Vibe Code Bench v1.1

51.48%

±4.64

50/108

Harvey's Legal Agent Benchmark

5.00%

±1.65

32/74

GPQA Diamond

85.61%

±2.41

48/138

LiveCodeBench

82.09%

±1.06

57/143

LegalBench

82.12%

±0.48

73/149

MMLU Pro

87.34%

±0.43

29/138

MMMU Pro

83.58%

±0.89

31/93

ProgramBench

0.50%

±0.50

28/62

SkillsBench

49.05%

±4.64

24/35

SWE-bench

77.40%

±1.87

35/88

Terminal-Bench 2.1

57.30%

±0.65

45/75

Proprietary BenchmarksAcademic BenchmarksIndustry Partners

View Hyperparameter Settings

Default Provider : Anthropic

Temperature: 1

Top P: Default

Top K: Default

Max Output Tokens: 128,000

Compute Effort: max

Compared with

Claude Sonnet 4.6 vs Grok 4.5 18 shared benchmarks Claude Sonnet 4.6 vs Qwen 3.8 Max 19 shared benchmarks Claude Sonnet 4.6 vs GLM 5.3 Flash 16 shared benchmarks Claude Sonnet 4.6 vs DeepSeek V4 Pro 0813 15 shared benchmarks Claude Sonnet 4.6 vs Gemini 3.5 Flash 19 shared benchmarks Claude Sonnet 4.6 vs GLM 5.2 14 shared benchmarks Claude Sonnet 4.6 vs DeepSeek V4 Flash 0731 15 shared benchmarks Claude Sonnet 4.6 vs Gemini 3.6 Flash 18 shared benchmarks

Updates

Feb 17, 2026

Claude Sonnet 4.6 - Anthropic's latest model

Anthropic’s latest model, Claude Sonnet 4.6, debuts at #1 on our Vals Index and #1 on the Vals Multimodal Index, evaluated across 17 benchmarks. Key takeaways:

Claude Sonnet 4.6 takes first place on both Finance Agent (63.3%) and Tax Eval v2 (77.1%), demonstrating strong domain expertise in finance and tax.
It takes first place on Terminal-Bench 2.0 (59.55%), beating out Claude Opus 4.5 (Thinking) and Claude Opus 4.6 (Thinking). It is also #3 on SWE-bench Verified (76.2%) and #2 on ProofBench (45.0%), showing competitive performance on agentic coding and formal mathematics.
On knowledge benchmarks, Sonnet 4.6 scores 92.3% on AIME, 87.3% on MMLU Pro (#5), and 85.6% on GPQA Diamond.
The model finishes top 10 on 13 of 17 benchmarks, with particularly strong results across finance and legal domains including #6 on CaseLaw and #7 on Mortgage Tax.