The LLM benchmark aggregator.

Claude Sonnet 4.6

Anthropic

Strongest in Computer use (#11 of 68, on 1 of its 2 boards), weakest in Secure code (#154 of 270). Above par in 18 of 25 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 4.8 and Claude Fable 5 and ahead of Qwen3.7 Max and Claude Haiku 5.5.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Professional55.7−6.7#25/1941/2

Finance57.4−5#24/1621/1

Legal55.6−10.3#42/1731/1

Medical54.6−11.9#48/1491/1

Human preference63.6−5.2#39/3541/1

Agents56.4−10#42/2983/6

Computer use61.8−6.3#11/681/2

Knowledge work55.6−15.1#62/2082/2

Knowledge53.2−9.3#57/1971/3

Academic knowledge57.6−6.3#29/1231/1

Reasoning53.8−11.6#59/1953/5

Science56.6−6.9#48/1221/1

Reasoning53.7−15.3#49/1471/3

Mathematics52.5−11.3#74/1522/2

Coding49.7−16.5#90/1792/6

Agentic coding49.7−17.2#88/1732/5

Safety52−9.4#131/3292/3

Harm refusal56.5−5#41/2921/2

Factual grounding48.3−22.4#65/1041/1

Fairness47.9−22.1#142/2921/2

Secure code50.1−16.1#154/2701/1

Core abilities46−19.3#133/1991/4

Language47.4−19.2#41/621/1

Instruction following41.5−32.3#48/621/1

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Claude Sonnet 4.6, left for the other.

NVIDIA Nemotron 3.5 Lightning 30B A3Bwon 15 · lost 0

MiMo V2.5won 15 · lost 0

Mistral Medium 3.5won 16 · lost 0

Inkling Smallwon 17 · lost 0

Llama 4 Scout Instructwon 19 · lost 1

Command Awon 19 · lost 1

Gemini 2.0 Flashwon 18 · lost 1

Mistral Large 3won 18 · lost 1

Gemini 3.5 Flash-Litewon 18 · lost 1

Llama 3.3 70B Instructwon 17 · lost 1

GLM-4.7won 17 · lost 1

Ling 3.0 Flashwon 15 · lost 1

Gemini 1.5 Prowon 15 · lost 1

Magistral Small 1.2won 14 · lost 1

Command R+won 14 · lost 1

GPT-4.1 nanowon 19 · lost 2

GPT-4o miniwon 18 · lost 2

GPT-5.4 miniwon 18 · lost 2

MiMo V2.5 Prowon 17 · lost 2

Kimi K2.5won 20 · lost 3

Llama 4 Maverick Instructwon 19 · lost 3

Grok 4.3won 19 · lost 3

Gemma 4 31Bwon 16 · lost 3

Claude 3.5 Haikuwon 15 · lost 3

Gemini 3.1 Flash-Litewon 15 · lost 3

Qwen3.6 Pluswon 15 · lost 3

NVIDIA Nemotron 3 Ultra 550B A55Bwon 19 · lost 4

Gemini 2.5 Flash-Litewon 14 · lost 3

Grok 4 Fastwon 14 · lost 3

DeepSeek V3won 18 · lost 4

MiniMax M2.1won 13 · lost 3

GLM-4.5won 13 · lost 3

GPT-4.1 miniwon 17 · lost 4

GPT-5.4 nanowon 17 · lost 4

DeepSeek R1won 17 · lost 4

Qwen3.5 Flashwon 12 · lost 3

Grok 4.1 Fastwon 16 · lost 4

GPT OSS 20Bwon 14 · lost 4

GPT OSS 120Bwon 17 · lost 5

Gemini 2.5 Prowon 17 · lost 5

MiniMax M2.5won 13 · lost 4

GLM-5won 13 · lost 4

GPT-4owon 16 · lost 5

Claude Haiku 4.5won 16 · lost 5

Claude 3.7 Sonnetwon 15 · lost 5

Grok 3won 15 · lost 5

GLM-5.1won 15 · lost 5

Inklingwon 18 · lost 6

DeepSeek V3.2won 16 · lost 6

GPT-6 Lunawon 16 · lost 6

Gemini 2.5 Flashwon 15 · lost 6

Mistral Large 4won 15 · lost 6

GLM-4.6won 15 · lost 6

MiMo V2.6 Flashwon 12 · lost 5

MiniMax M2.7won 14 · lost 6

Claude 3.5 Sonnetwon 11 · lost 5

GPT-5 miniwon 15 · lost 7

Grok 3 Miniwon 12 · lost 6

Kimi K2won 14 · lost 7

MiniMax M3won 16 · lost 8

Kimi K2.6won 16 · lost 8

Claude Sonnet 4won 15 · lost 8

DeepSeek V4 Flashwon 15 · lost 8

Claude Sonnet 4.5won 15 · lost 8

Claude Opus 4won 13 · lost 7

Qwen3.8 27Bwon 13 · lost 7

Kimi K2 Thinkingwon 11 · lost 6

GPT-5 nanowon 12 · lost 7

O4 Miniwon 13 · lost 8

Qwen3 235B A22B Instructwon 13 · lost 8

O1won 11 · lost 7

GLM-5.2won 14 · lost 9

O3 Miniwon 12 · lost 8

DeepSeek V4 Prowon 14 · lost 10

GPT-4.1won 12 · lost 9

O3won 12 · lost 9

GPT-5.1won 12 · lost 9

GPT-5.6 Lunawon 13 · lost 10

Claude Haiku 5.5won 10 · lost 8

Qwen3.7 Maxwon 12 · lost 11

Gemini 3 Flashwon 8 · lost 8

Grok 4won 10 · lost 10

GLM-5.3 Flashwon 12 · lost 12

GPT-5won 10 · lost 12

Claude Opus 4.5won 9 · lost 11

Gemini 3.6 Flashwon 10 · lost 13

Grok 4.7won 9 · lost 12

Gemini 3.1 Prowon 9 · lost 12

Muse Sparkwon 6 · lost 9

Gemini 3 Prowon 8 · lost 12

Grok 4.5won 9 · lost 14

Gemini 3.5 Flashwon 9 · lost 14

MiMo V2.6 Prowon 7 · lost 12

GPT-5.2won 8 · lost 15

Claude Sonnet 5won 8 · lost 15

Grok 4.6won 8 · lost 16

GPT-5.6 Terrawon 8 · lost 16

GPT-6 Solwon 7 · lost 15

GLM-5.3won 6 · lost 13

GPT-5.4won 7 · lost 17

Qwen3.8 Maxwon 7 · lost 17

Claude Sonnet 5.5won 5 · lost 13

Muse Spark 1.1won 6 · lost 17

Gemini 3.8 Flashwon 6 · lost 18

GPT-6 Astrawon 4 · lost 16

DeepSeek V4.1 Flashwon 4 · lost 16

Muse Spark 1.3won 3 · lost 13

Claude Opus 4.6won 4 · lost 18

Gemini 4 Argonwon 3 · lost 14

GPT-6.1 Solwon 3 · lost 15

GPT-5.6 Solwon 4 · lost 20

Gemini 3.7 Flashwon 3 · lost 16

Claude Opus 5.5won 3 · lost 20

Muse Wonderwon 2 · lost 14

Muse Spark 1.2won 2 · lost 16

Kimi K3won 2 · lost 18

GPT-5.5won 2 · lost 22

Claude Opus 4.7won 2 · lost 22

Claude Fable 5.1won 1 · lost 19

Claude Opus 5won 1 · lost 23

Claude Fable 5won 0 · lost 20

Claude Opus 4.8won 0 · lost 23

§ 3 · Sources

Where the numbers come from

9 publications, 32 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-11-10

LMArena Text 1472

GDPval-AA ↗1 measureread 2026-11-10

GDPval-AA 1241

AA-Briefcase ↗1 measureread 2026-11-10

AA-Briefcase 1062

ARC-AGI-2 ↗1 measureread 2026-11-10

ARC-AGI-2 58.3%

LiveBench 1.0 ↗8 measuresread 2026-11-10

LiveBench 1.0 73LiveBench 1.0 · Reasoning 84.8LiveBench 1.0 · Coding 79.3LiveBench 1.0 · Agentic Coding 42.6LiveBench 1.0 · Mathematics 87LiveBench 1.0 · Data Analysis 77.9LiveBench 1.0 · Language 76.1LiveBench 1.0 · Instruction Following 63.2

OSWorld ↗1 measureread 2026-11-10

OSWorld 72.1%

Vals.ai ↗14 measuresread 2026-11-10

Vals · Legal Research Bench 38.46%Vals · CaseLaw 63.99%Vals · LegalBench 82.12%Vals · Finance Agent 51.03%Vals · CorpFin 65.31%Vals · TaxEval 77.11%Vals · MortgageTax 67.73%Vals · MedQA 92.06%Vals · SWE-bench Verified 77.4% (Mini-SWE-agent)Vals · Vibe Code Bench 55.77% (Claude Code)Vals · Code Migration 39.89%Vals · GPQA Diamond 85.61%Vals · MMLU Pro 87.34%Vals · AIME 92.29%

Enkrypt AI Safety Leaderboard ↗4 measuresread 2026-11-10

Enkrypt · Harmful content risk 0.6%Enkrypt · CBRN risk 15.3%Enkrypt · Bias risk 81.1%Enkrypt · Insecure code risk 32.9%

Vectara Hallucination Leaderboard ↗1 measureread 2026-11-10

Vectara · Factual consistency 89.4%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=claude-sonnet-4-6)](https://publicai.io/model-index/m/claude-sonnet-4-6)