Anthropic · proprietary

Claude Sonnet 4.6

Claude Sonnet 4.6 by Anthropic appears in 2 sources with Reasoning at 52.16. Best read for Code quality, Coding, High intelligence.

CreatorAnthropic

Release date2026-02-17

Knowledge cutoffNot published

Context200K tokens

Input price$3/M tokens

Output price$15/M tokens

Modalitytext + vision

CountryUS

Metrics

All source-backed metrics

Open in leaderboard

LLM Stats Rank8ranking · source_rank

Reasoning52.16reasoning · index_reasoning

Math44.18math · index_math

Coding36.51coding · index_code

Research24.59research · index_search

Writing35.55writing · index_communication

Vision33.25multimodal · index_vision

Tool calling28.76tool_calling · index_tool_calling

Long context26.42long_context · index_long_context

Finance37.5domain · index_finance

Legal35.41domain · index_legal

Healthcare14.15domain · index_healthcare

GPQA89.9 %reasoning · gpqa_score

SWE-bench Verified79.6 %coding · swe_bench_verified_score

Code Arena1,700.63 %coding · coding_arena_score

Humanity Last Exam49 %reasoning · hle_score

ARC-AGI v258.3 %reasoning · arc_agi_v2_score

MMMLU89.3 %reasoning · mmmlu_score

MMMU-Pro75.6 %multimodal · mmmu_pro_score

OSWorld72.5 %agent · osworld_score

BrowseComp74.7 %research · browsecomp_score

MCP Atlas61.3 %tool_calling · mcp_atlas_score

Context200,000 tokenscontext · context

Speed120.6 c/sperformance · throughput

Latency8,045.6 msperformance · latency

Input price3 $/Mpricing · input_price

Output price15 $/Mpricing · output_price

Arena Rating1,453.06metric

Arena Rank20metric

Vote Count19,224metric

Evidence

Citations and source overlap

Arena AILM Arena text leaderboard (overall) mirrored in the public leaderboard dataset.Retrieved 2026-05-20 LLM StatsPublic full LLM leaderboard advanced explorer snapshot, including ranking, benchmark indexes, pricing, speed, context, license, and model metadata.Retrieved 2026-06-01

FAQ

How should I read this profile?

Treat this as a source-backed model dossier, not an EvalKit-run verification. The public values are replicated from linked sources and kept source-scoped.

Is Claude Sonnet 4.6 verified by EvalKit?

No. EvalKit currently shows 0 verified rows until real run evidence exists.

Why can metrics disagree?

Different sources test different tasks, dates, prompts, and aggregation methods. EvalKit keeps those differences visible instead of merging them into a fake universal score.