Request a pilot
‹ PublicAI Index

The LLM benchmark aggregator.

Claude Sonnet 4

Anthropic

Strongest in Toxicity avoidance (#12 of 272), weakest in Agents (#127 of 268). Above par in 17 of 28 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5 and Claude Fable 5.1 and ahead of Kimi K2.6 and Claude 3.5 Sonnet.

§ 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.

Core abilities56.9#27/204
General intelligence59.9#28/204
Safety56.6#28/337
Toxicity avoidance58.7#12/272
Jailbreak resistance61.8#23/272
Fairness62.1#31/300
Safe-prompt compliance55.5#31/82
Factual grounding49.2#58/101
Secure code59.3#70/274
Harm refusal54.2#84/300
Coding52.7#54/165
Agentic coding54.4#44/157
Code generation50#46/77
Knowledge52.7#73/138
Academic knowledge53.3#67/123
Professional48.7#94/168
Legal52.9#54/151
Finance48.7#95/151
Medical46.1#107/140
Reasoning48.6#97/178
Mathematics52.7#69/140
Science49.2#79/122
Reasoning45.1#82/140
Human preference57.2#122/342
Human preference57.2#122/342
Agents49.3#127/268
Computer use48.9#18/37
Knowledge work49.4#87/178

§ 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, left for the other.

Gemini 1.5 Flashwon 17 · lost 0
GPT-4o Miniwon 25 · lost 0
GPT-4.1 Nanowon 24 · lost 1
Llama 4 Scout Instructwon 23 · lost 1
Gemini 2.0 Flashwon 22 · lost 1
Command R Pluswon 17 · lost 1
Mistral Smallwon 18 · lost 2
Gemini 1.5 Prowon 17 · lost 2
Mistral Medium 3.5won 15 · lost 2
Gemini 2.5 Flash Litewon 16 · lost 3
Ling 3.0 Flashwon 15 · lost 3
Command Awon 19 · lost 4
DeepSeek V3won 22 · lost 5
Claude 3.5 Haikuwon 16 · lost 4
GPT-4.1 Miniwon 20 · lost 5
Mistral Largewon 14 · lost 4
GPT OSS 20Bwon 17 · lost 5
Mistral Large 3won 15 · lost 5
O3 Miniwon 18 · lost 6
GPT OSS 120Bwon 19 · lost 7
GPT-4owon 16 · lost 7
GPT-5 Nanowon 16 · lost 7
Claude Haiku 4.5won 18 · lost 8
DeepSeek R1won 18 · lost 8
Grok 4.3won 17 · lost 8
GPT-5.4 Nanowon 14 · lost 7
Grok 3won 14 · lost 7
Kimi K2won 16 · lost 9
Gemma 4 31Bwon 14 · lost 8
GLM-4.5won 12 · lost 7
GLM-4.6won 15 · lost 9
Mimo V2.5 Prowon 14 · lost 9
Kimi K2 Thinkingwon 12 · lost 8
O4 Miniwon 15 · lost 11
MiniMax M3won 15 · lost 11
Claude 3.7 Sonnetwon 13 · lost 10
GPT-4.1won 14 · lost 11
DeepSeek V3.2won 14 · lost 11
GPT-5.4 Miniwon 11 · lost 9
Gemini 2.5 Flashwon 14 · lost 12
Gemini 3.1 Flash Litewon 10 · lost 9
Claude 3.5 Sonnetwon 11 · lost 10
Kimi K2.6won 14 · lost 13
Grok 4 Fastwon 9 · lost 9
Grok 4.1 Fastwon 11 · lost 11
GPT-5 Miniwon 13 · lost 13
Qwen3 235B A22B Instructwon 12 · lost 13
Qwen3.7 Maxwon 12 · lost 13
Gemini 3.6 Flashwon 12 · lost 13
Gemini 3.1 Prowon 10 · lost 11
Qwen3.6 Pluswon 9 · lost 10
Gemini 3.5 Flash Litewon 9 · lost 10
Gemini 2.5 Prowon 12 · lost 15
DeepSeek V4 Flashwon 11 · lost 14
GLM-5.3 Flashwon 11 · lost 14
Gemini 3.5 Flashwon 11 · lost 14
GPT-6 Lunawon 9 · lost 13
GLM-5.1won 9 · lost 13
Inklingwon 10 · lost 15
Gemini 3.8 Flashwon 10 · lost 15
O1won 8 · lost 13
MiniMax M2.7won 8 · lost 13
Grok 4won 9 · lost 15
Claude Sonnet 4.5won 10 · lost 17
GPT-5.6 Lunawon 9 · lost 16
Claude Sonnet 4.6won 9 · lost 16
Kimi K2.5won 9 · lost 17
DeepSeek V4 Prowon 9 · lost 17
GPT-5.2won 8 · lost 17
Grok 4.5won 8 · lost 17
GLM-5.2won 8 · lost 17
Gemini 3 Prowon 7 · lost 15
Claude Opus 4.6won 7 · lost 15
Qwen3.8 27Bwon 6 · lost 13
Mimo V2.6 Prowon 6 · lost 13
GPT-5.4won 8 · lost 18
GPT-6 Solwon 7 · lost 16
Claude Opus 4.5won 6 · lost 14
Claude Opus 4won 7 · lost 18
GPT-5.6 Terrawon 7 · lost 18
GPT-5.1won 6 · lost 18
Qwen3.8 Maxwon 6 · lost 19
GPT-5won 6 · lost 20
Muse Spark 1.1won 6 · lost 20
Claude Opus 4.7won 6 · lost 20
GLM-5won 4 · lost 14
O3won 5 · lost 21
GPT-5.6 Solwon 5 · lost 21
Claude Opus 5won 5 · lost 21
Grok 4.7won 4 · lost 18
Muse Spark 1.2won 3 · lost 15
Grok 4.6won 4 · lost 21
Claude Opus 4.8won 4 · lost 21
GLM-5.3won 3 · lost 16
GPT-5.5won 4 · lost 22
DeepSeek V4.1 Flashwon 2 · lost 16
Gemini 3.7 Flashwon 2 · lost 17
Claude Opus 5.5won 2 · lost 20
Claude Sonnet 5won 2 · lost 23
GPT-6 Astrawon 1 · lost 17
Kimi K3won 1 · lost 18
Claude Fable 5.1won 0 · lost 19
Claude Fable 5won 0 · lost 20

§ 3 · Sources

Where the numbers come from

12 publications, 30 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-09-26
LMArena Text 1402
GDPval-AA ↗1 measureread 2026-09-26
GDPval-AA 680
ARC-AGI-2 ↗1 measureread 2026-09-26
ARC-AGI-2 5.9%
LiveCodeBench ↗1 measureread 2026-09-26
LiveCodeBench 55.9%
Aider polyglot ↗1 measureread 2026-09-26
Aider polyglot 61.3%
OSWorld ↗1 measureread 2026-09-26
OSWorld 43.9%
Kagi LLM Benchmark ↗1 measureread 2026-09-26
Kagi LLM Benchmark 73%
SimpleBench ↗1 measureread 2026-09-26
SimpleBench 45.5%
Vals.ai ↗10 measuresread 2026-09-26
Vals · LegalBench 82.06%Vals · CorpFin 61.23%Vals · TaxEval 72%Vals · MortgageTax 49.88%Vals · MedQA 92.71%Vals · MedCode 34.96%Vals · MedScribe 69.35%Vals · GPQA Diamond 75%Vals · MMLU Pro 83.86%Vals · AIME 76.25%
HELM Safety ↗5 measuresread 2026-09-26
HELM Safety · HarmBench 98%HELM Safety · SimpleSafetyTests 100%HELM Safety · Anthropic Red Team 99.2%HELM Safety · BBQ 96.6%HELM Safety · XSTest 96.6%
Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26
Enkrypt · Jailbreak risk 2.8%Enkrypt · Harmful content risk 0.6%Enkrypt · CBRN risk 37%Enkrypt · Toxicity risk 0.2%Enkrypt · Bias risk 57.1%Enkrypt · Insecure code risk 14.7%
Vectara · Factual consistency 89.7%

Badge

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