Inference
EN
Anthropic

Claude Sonnet 4.6

Claude Sonnet 4.6 is a standard model from Anthropic. 42 benchmarks count toward its index, in 8 categories.

  • Closed weights
  • Standard
  • Input: text · image · file
  • Released 2026-02
RankedEnough results for an index.
IndexAll categories, out of 100.57.0±2.5Median 51.0
CoverageIndex weight that has results.100%
SpeedOutput tokens per second.44/sMedian 67/s
First tokenSeconds to the first output token.0.6 sMedian 1.3 s
Input / 1MUS dollars per 1M input tokens.$3.00batch $1.50Median $0.60
Output / 1MUS dollars per 1M output tokens.$15.00batch $7.50 Batch price per 1M output tokens.Median $2.55
ContextMaximum tokens in one request.1MMedian 262K
EloLMArena rating and rank.1458±4 #43Median 1422

Details

From OpenRouter
API IDThe model ID on OpenRouter.
anthropic/claude-sonnet-4.6
Max outputMaximum output tokens per request.
128K
Knowledge cutoffLast date of training data.
—
ToolsTool calls on OpenRouter.
Yes
JSON outputOutput follows a given JSON schema.
Yes
Cache read / 1MUS dollars per 1M cached tokens.
$0.30
WeightsLink to the published weights.
—
RetiresDate OpenRouter removes the model.
—

Where to run

APIs that serve this model.
9 providers
ProviderTypeMaker, cloud, host or router.InputOutputContext
Amazon BedrockCloud$3.00$15.001M
DatabricksCloud$3.00$15.001M
DigitalOceanCloud$3.00$15.00200K
GMI CloudHost$3.00$15.00410K
Google Vertex AICloud$3.00$15.001M
Microsoft AzureCloud$3.00$15.001M
OpenRouterRouter$3.00$15.001M
SAP AI CoreCloud$3.00$15.001M
Snowflake CortexCloud$3.00$15.001M

US dollars per 1M tokens, list price. From models.dev (opens in a new tab) (MIT).

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.13 counted59.769 / 218
CodingCode writing and repair.8 counted58.976 / 209
ReasoningLogic problems and puzzles.7 counted56.476 / 227
MultimodalTasks with images and text.2 counted53.958 / 84
KnowledgeFacts and expert knowledge.6 counted57.278 / 268
MultilingualTasks in many languages.1 counted56.6—
InstructionTasks with strict prompt rules.1 counted27.5—
MathMath problems.4 counted54.168 / 168

Results

42 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
SuperGPQA: Scaling LLM Evaluation Across 285 Graduate DisciplinesKnowledgeAbove fitFar above this model's index.95.0%1 / 22
EuroEval FrenchMultilingual68.5% ±0.412 / 178
EuroEval PortugueseMultilingual67.0% ±0.513 / 182
EuroEval SpanishMultilingual58.5% ±0.518 / 176
EuroEval PolishMultilingual59.9% ±0.522 / 209
Terminal-Bench 2.0 · Vals AIAgentic59.6% ±5.27 / 66
EuroEval GermanMultilingual54.8% ±0.521 / 191
EuroEval SwedishMultilingual64.5% ±0.722 / 195
EuroEval ItalianMultilingual65.3% ±0.520 / 174
EuroEval DutchMultilingual59.7% ±0.426 / 175
Gert Labs Composite Game BenchmarkAgentic62.9%8 / 52
Software Engineering Benchmark VerifiedCoding79.6%23 / 114
AIMEMath92.3% ±0.420 / 95
MMLU ProKnowledge87.3% ±0.429 / 135
Graduate-Level Google-Proof Q&AKnowledge89.9%31 / 141
FrontierMath-2025-02-28-PrivateMath32.4% ±2.816 / 69
Humanity's Last ExamKnowledge49.0%22 / 77
MMMU ProMultimodal83.6% ±0.930 / 92
GPQA Diamond · Vals AIKnowledge85.6% ±2.445 / 135
Claw-EvalAgentic67.8%14 / 42
ARC-AGI-2 (semi-private)Reasoning58.3%45 / 131
Artificial Analysis Omniscience AccuracyKnowledge38.6%58 / 164
ARC-AGI-1 (semi-private)Reasoning86.0%48 / 126
FrontierMath-Tier-4-2025-07-01-PrivateMath8.3% ±4.021 / 55
Code MigrationCoding39.9% ±4.130 / 77
LiveCodeBenchCoding82.1% ±1.156 / 142
SWE-benchCoding77.4% ±1.935 / 87
LiveBench ReasoningReasoning86.4%33 / 80
GPQA diamond · Epoch AIKnowledge78.8% ±2.993 / 217
React Native EvalsCoding80.6%7 / 16
SWE-Bench verified · Epoch AICoding75.2% ±2.014 / 32
LiveBench Data AnalysisReasoning76.1%37 / 80
LiveBench Browser UseAgentic68.0%23 / 49
τ²-Bench Tool-Agent-User EvaluationAgentic79.5%70 / 149
MazeBench: visual spatial reasoning in a 3D open worldAgentic0.0%23 / 48
Terminal-Bench 2.0 · BenchLMAgentic59.1%20 / 41
OTIS Mock AIME 2024-2025Math71.1% ±6.898 / 200
Artificial Analysis GPQA DiamondKnowledge79.9%100 / 199
Vibe Code Bench v1.1Coding51.5% ±4.665 / 123
Massive Multitask Language Understanding ProfessionalKnowledge79.2%40 / 75
JobBenchAgentic36.9%20 / 36
Artificial Analysis Long Context ReasoningReasoning68.3%122 / 217
Critical Physics TasksReasoning0.9%124 / 214
Terminal-Bench 2.1 · Vals AIAgentic57.3% ±0.644 / 75
Artificial Analysis Humanity's Last ExamKnowledge13.3%131 / 222
OSWorld-VerifiedAgentic72.1%22 / 37
Chess PuzzlesReasoning23.0% ±4.256 / 90
Artificial Analysis MMMU-ProMultimodal70.6%73 / 113
SkillsBenchCoding49.1% ±4.624 / 35
Artificial Analysis IFBenchInstruction41.2%101 / 146
LiveBench DesignAgentic71.8%36 / 52
Mystery Game PuzzlesReasoning16.0% ±3.756 / 80
SimpleQA VerifiedKnowledge32.8% ±1.565 / 91
LiveBench Long AgenticAgentic23.1%38 / 51
LiveBench ResearchAgentic63.6%41 / 52
LiveBench Coding 2.0Coding66.6%42 / 52
CyberGymAgentic65.2%25 / 29
CharXiv ReasoningMultimodal77.4%43 / 49
OSWorld 2.0Agentic8.3%22 / 25
FrontierCode 1.1 MainCoding24.3%17 / 17
ApprenticeBench: end-to-end computer use, continual learning, and long-horizon agency on a real accounts-payable jobAgentic2.0%18 / 18
TaxEval v2Finance77.1% ±0.84 / 143
Finance Agent v1.1Finance63.3% ±2.82 / 51
FrontierMath-Tier-4-2025-07-01-PublicMath0.0%3 / 36
FrontierMath-2025-02-28-PublicMath80.0% ±13.36 / 64
CorpFin v2Finance65.3% ±0.928 / 131
MortgageTaxFinance67.7% ±0.922 / 97
ARC-AGI-1 (public eval)Reasoning95.8%30 / 116
CaseLaw v2Legal64.0% ±0.614 / 53
ARC-AGI-2 (public eval)Reasoning62.4%35 / 122
ProgramBenchCoding0.5% ±0.519 / 65
Public Benefits Bench v1Public benefits58.5% ±1.34 / 13
MedQAHealthcare92.1% ±0.235 / 95
Legal Research BenchLegal38.5% ±3.439 / 90
EMBFinance60.2% ±3.334 / 75
LegalBenchLegal82.1% ±0.575 / 149
Finance Agent (v2)Finance51.0% ±0.341 / 79
Tax Agent BenchFinance25.8% ±3.243 / 82
Public Benefits Bench v1.1Public benefits62.4% ±1.327 / 50
SWE-RebenchCoding60.7%—
cursorBench31Coding48.8%—

Sources

BenchLMCC BY-NC 4.0OpenRouterTerms restrict reuseVals AINo licence statedLiveBenchNo licence statedARC Prize FoundationTerms restrict reuseEpoch AICC BYEuroEvalMITmodels.devMIT

History

5 releases
Index58.1 → 57.0
23 Oct 202612 Nov 2026
Output / 1M$15.00 → $15.00
23 Oct 202612 Nov 2026

Questions

How much does Claude Sonnet 4.6 cost?
$3.00 per 1M input tokens, $15.00 per 1M output tokens.
How good is Claude Sonnet 4.6?
Its index is 57.0 out of 100. That is place 92 of 283 ranked models.
What is Claude Sonnet 4.6 best at?
Knowledge: 57.2 out of 100, place 78 of 268.
How fast is Claude Sonnet 4.6?
44 output tokens per second, 0.6 s to the first token.
What is the context window of Claude Sonnet 4.6?
1M tokens.
Is Claude Sonnet 4.6 open source?
No. Its weights are not published.
When was Claude Sonnet 4.6 released?
February 2026.

Badge

Claude Sonnet 4.6 on Inference360: place 92 of 283, index 57.0
HTML
<a href="https://inference360.ai/models/claude-sonnet-4-6"><img src="https://inference360.ai/badges/claude-sonnet-4-6.svg" alt="Claude Sonnet 4.6 on Inference360: place 92 of 283, index 57.0" height="20"></a>
Markdown
[![Claude Sonnet 4.6 on Inference360: place 92 of 283, index 57.0](https://inference360.ai/badges/claude-sonnet-4-6.svg)](https://inference360.ai/models/claude-sonnet-4-6)

Compare with

Qwen 3.6 Max (preview)57.2MiMo-V2.5-Pro56.3Claude Opus 4.5 Thinking56.3Claude Opus 4.658.0Claude Opus 4.7 (Adaptive)63.2

Lower price

dots3-note Preview63.4 · FreeApodex 1.1 Mini55.7 · FreeGLM-5.3-Flash63.1 · $0.078/1M

More from Anthropic

All models
Claude Opus 4.6 (Adaptive)60.0Claude Opus 4.658.0Claude Mythos PreviewUnrankedClaude Opus 4.5 Thinking56.3Claude Opus 4.552.0