← All models

Anthropic · proprietary

Claude Opus 4.6

Claude Opus 4.6 by Anthropic appears in 3 sources with Reasoning at 59.45. Best read for Code quality, Coding, High intelligence.

CreatorAnthropic
Release date2026-02-05
Knowledge cutoffNot published
Context1M tokens
Input price$5/M tokens
Output price$25/M tokens
Modalitytext + vision
CountryUS

Metrics

All source-backed metrics

Open in leaderboard
LLM Stats Rank1ranking · source_rank
Reasoning59.45reasoning · index_reasoning
Math52.7math · index_math
Coding43.61coding · index_code
Research36.7research · index_search
Writing44.61writing · index_communication
Vision35.65multimodal · index_vision
Tool calling33.83tool_calling · index_tool_calling
Long context36.25long_context · index_long_context
Finance35.14domain · index_finance
Legal33.59domain · index_legal
Healthcare13.93domain · index_healthcare
GPQA91.3 %reasoning · gpqa_score
AIME 202599.79 %math · aime_2025_score
SWE-bench Verified80.8 %coding · swe_bench_verified_score
Code Arena2,125.34 %coding · coding_arena_score
Humanity Last Exam53.1 %reasoning · hle_score
ARC-AGI v268.8 %reasoning · arc_agi_v2_score
MMMLU91.1 %reasoning · mmmlu_score
MMMU-Pro77.3 %multimodal · mmmu_pro_score
OSWorld72.7 %agent · osworld_score
BrowseComp84 %research · browsecomp_score
MCP Atlas62.7 %tool_calling · mcp_atlas_score
MRCR v293 %long_context · mrcr_v2_score
CharXiv-R77.4 %multimodal · charxiv_r_score
Context1,000,000 tokenscontext · context
Speed50.56 c/sperformance · throughput
Latency3,118.21 msperformance · latency
Input price5 $/Mpricing · input_price
Output price25 $/Mpricing · output_price
Arena Rating1,497.8metric
Arena Rank2metric
Vote Count27,338metric
AIME 202599.8 %metric
Humanity's Last Exam40 %metric
SWE-Bench Verified80.8 %metric

Evidence

Citations and source overlap

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 Opus 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.