Request a pilot
‹ PublicAI Index

The LLM benchmark aggregator.

Claude 3 Sonnet

Anthropic

Strongest in Harm refusal (#6 of 300), weakest in Human preference (#236 of 342). Above par in 6 of 9 scopes. Among the models it meets almost everywhere, it finishes behind Claude 3.5 Sonnet and GPT-5 Nano and ahead of Gemini 2.5 Pro and DeepSeek R1.

§ 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.

Safety55.5#42/337
Harm refusal60.2#6/300
Jailbreak resistance62.8#15/272
Toxicity avoidance58.4#18/272
Safe-prompt compliance31.5#76/82
Secure code56.7#103/274
Fairness50.4#109/300
Human preference45.5#236/342
Human preference45.5#236/342

§ 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 3 Sonnet, left for the other.

OLMo 7B Instruct Hfwon 6 · lost 0
SmolLM 1.7B Instructwon 6 · lost 0
Qwen2.5 7B Instructwon 6 · lost 0
H2o Danube3 500M Chatwon 6 · lost 0
LFM2 350Mwon 6 · lost 0
Laguna Xs 2.1won 6 · lost 0
RakutenAI 7B Instructwon 6 · lost 0
GLM-4 9B Chatwon 6 · lost 0
Qwen2.5 0.5B Instructwon 6 · lost 0
Qwen2 7B Instructwon 6 · lost 0
Jamba Mini 1.7won 6 · lost 0
Magistral Smallwon 6 · lost 0
Aya 23 8Bwon 6 · lost 0
Laguna S 2.1won 6 · lost 0
H2o Danube3 4B Chatwon 6 · lost 0
Ling 3.0 Flash VLwon 6 · lost 0
Jamba 1.6 Miniwon 6 · lost 0
Qwen3 8Bwon 6 · lost 0
Open Mixtral 8x7bwon 6 · lost 0
Phi-4 Mini Instructwon 6 · lost 0
Qwen2.5 3B Instructwon 6 · lost 0
Aya 23 35Bwon 6 · lost 0
DeepSeek V3p1won 6 · lost 0
Jamba 1.6 Largewon 6 · lost 0
Yi 1.5 9B Chatwon 6 · lost 0
Jamba Instructwon 6 · lost 0
SeaLLMs V3 7B Chatwon 6 · lost 0
Qwen2.5 7B Instruct 1Mwon 6 · lost 0
DeepSeek Reasonerwon 6 · lost 0
Qwen2.5 1.5B Instructwon 6 · lost 0
WizardLM 2.8x22Bwon 6 · lost 0
Hunyuan A13B Instructwon 6 · lost 0
Open Mixtral 8x22bwon 6 · lost 0
Apertus 8B Instructwon 6 · lost 0
DeepSeek Chatwon 6 · lost 0
Qwen2.5 32B Instructwon 6 · lost 0
LongWriter Glm4 9Bwon 6 · lost 0
Sarvam 30Bwon 6 · lost 0
Qwen2.5 14B Instructwon 6 · lost 0
Llama 3 70B Chat HFwon 6 · lost 0
Apertus 70B Instructwon 6 · lost 0
Granite 4.0 H Microwon 6 · lost 0
Granite 4.0 H Tinywon 6 · lost 0
Phi 3.5 Mini Instructwon 6 · lost 0
Krutrim 2 Instructwon 6 · lost 0
EAI SageAlign 8B LoRAwon 6 · lost 0
Llama 3 8B Chat Hfwon 6 · lost 0
Llama 3.2 Vision (11B)won 6 · lost 0
Llama 2 70B Chat Hfwon 6 · lost 0
Llama 2 13B Chat Hfwon 6 · lost 0
Granite 4.0 H Smallwon 6 · lost 0
Llama 2 7B Chat Hfwon 6 · lost 0
OLMoE 1B-7B Instructwon 7 · lost 0
Gemma 2B Itwon 8 · lost 0
Llama 3.2 1B Instructwon 8 · lost 0
Smollm2 1.7B Instructwon 8 · lost 0
Zephyr 7B Betawon 8 · lost 0
Gemma 7B Itwon 8 · lost 0
Llama 3.2 3B Instructwon 8 · lost 0
Starling Lm 7B Betawon 8 · lost 0
Qwen1.5 14B Chatwon 8 · lost 0
Llama 3.1 8B Instructwon 8 · lost 0
Llama 3.1 Tulu 3 8Bwon 8 · lost 0
Nova Microwon 8 · lost 0
Nova Litewon 8 · lost 0
Gemma 2 9B Itwon 8 · lost 0
DBRX Instructwon 9 · lost 0
Claude 3 Haikuwon 9 · lost 0
DeepSeek LLM 67B Chatwon 8 · lost 1
GPT-3.5 Turbowon 8 · lost 1
Llama 3 8B Instructwon 8 · lost 1
Command R Pluswon 8 · lost 1
Phi 3 Mini 4k Instructwon 7 · lost 1
C4ai Aya Expanse 8Bwon 7 · lost 1
Jamba 1.5 Miniwon 7 · lost 1
C4ai Aya Expanse 32Bwon 7 · lost 1
GPT-4won 7 · lost 1
OLMo 2 7B Instructwon 6 · lost 1
IBM Granite 4.0 Microwon 6 · lost 1
RakutenAI 7B Chatwon 5 · lost 1
PowerMoE 3Bwon 5 · lost 1
Phi 2won 5 · lost 1
PowerLM 3Bwon 5 · lost 1
SmolLM2 360M Instructwon 5 · lost 1
SEA LION V1 7B ITwon 5 · lost 1
Grok Build 0.1won 5 · lost 1
Command R7Bwon 5 · lost 1
LFM2 700Mwon 5 · lost 1
Jamba Large 1.7won 5 · lost 1
SmolLM 360M Instructwon 5 · lost 1
NexusRaven V2 13Bwon 5 · lost 1
Ling 3.0 Flashwon 5 · lost 1
Open Codestral Mambawon 5 · lost 1
LFM2 1.2Bwon 5 · lost 1
Gemma 4 E2Bwon 5 · lost 1
Gemma 4 E4Bwon 5 · lost 1
Sarvam 1won 5 · lost 1
Komodo 7B Basewon 5 · lost 1
Kimi K2.7 Codewon 5 · lost 1
Gemma 4 12Bwon 5 · lost 1
Fuguwon 5 · lost 1
Qwen2 VL 72Bwon 5 · lost 1
GLM-5 Turbowon 5 · lost 1
SeaLLM 7B V2won 5 · lost 1
K2 Chatwon 5 · lost 1
Phi 3.5 MoE Instructwon 5 · lost 1
Titan Text Express V1won 5 · lost 1
Internlm2 Chat 20Bwon 5 · lost 1
Llama 3 8B Instruct RRwon 5 · lost 1
Sarvam 105Bwon 5 · lost 1
Qwen1.5 72B Chatwon 5 · lost 1
Command Rwon 5 · lost 1
Olmo 2 32B Instructwon 5 · lost 1
Llama 3 70B Instructwon 7 · lost 2
Gemma 2 2B Itwon 6 · lost 2
Phi 4won 6 · lost 2
Jamba 1.5 Largewon 6 · lost 2
Nova Prowon 6 · lost 2
Gemma 3 4B Itwon 6 · lost 2
Magistral Mediumwon 6 · lost 2
QwQ 32Bwon 6 · lost 2
Gemma 3 12B Itwon 6 · lost 2
Command Awon 6 · lost 2
Mistral Smallwon 6 · lost 2
Gemma 3 27B Itwon 6 · lost 2
Qwen3 30B A3B Instructwon 6 · lost 2
Grok 3won 6 · lost 2
GLM-4.6won 6 · lost 2
Grok 4.1 Fastwon 6 · lost 2
DeepSeek V4 Flashwon 6 · lost 2
Grok 4.3won 6 · lost 2
Mimo V2.6 Flashwon 6 · lost 2
DeepSeek V4 Prowon 6 · lost 2
Mimo V2.6 Prowon 6 · lost 2
Gemini 3.6 Flashwon 6 · lost 2
GPT-5.6 Astrawon 4 · lost 2
Kimi K2 Thinkingwon 4 · lost 2
GPT OSS Safeguard 20Bwon 4 · lost 2
Fugu Ultra V2won 4 · lost 2
Nova Premierwon 4 · lost 2
Claude Sonnet 4.6won 4 · lost 2
Claude Opus 4.6won 4 · lost 2
Qwen2 72B Instructwon 6 · lost 3
Llama 4 Scout Instructwon 6 · lost 3
Gemini 1.5 Prowon 6 · lost 3
Gemini 2.0 Flash Litewon 6 · lost 3
Gemini 2.0 Flashwon 6 · lost 3
Grok 4won 6 · lost 3
Qwen2.5 72B Instructwon 5 · lost 3
Mistral Largewon 5 · lost 3
Qwen Maxwon 5 · lost 3
GLM-4.5won 5 · lost 3
Gemma 4 26B A4Bwon 5 · lost 3
MiniMax M3won 5 · lost 3
Kimi K2.5won 5 · lost 3
Gemma 4 31Bwon 5 · lost 3
GPT-5.6 Lunawon 5 · lost 3
GLM-5.3 Flashwon 5 · lost 3
GPT-5.4won 5 · lost 3
GLM-5.2won 5 · lost 3
Gemini 3.5 Flashwon 5 · lost 3
Qwen3.8 Maxwon 5 · lost 3
Gemini 1.5 Flashwon 5 · lost 4
GPT-4o Miniwon 5 · lost 4
GPT-4.1 Nanowon 5 · lost 4
GPT-4.1 Miniwon 5 · lost 4
Claude 3.7 Sonnetwon 5 · lost 4
DeepSeek V3won 5 · lost 4
Claude Haiku 4.5won 5 · lost 4
DeepSeek R1won 5 · lost 4
Gemini 2.5 Prowon 5 · lost 4
Cogito 671B V2 P1won 3 · lost 3
Meta SecAlign 8Bwon 3 · lost 3
Qwen3.5 35B A3Bwon 4 · lost 4
Qwen3.5 27Bwon 4 · lost 4
Qwen3.5 122B A10Bwon 4 · lost 4
DeepSeek V3.2won 4 · lost 4
GPT-6 Lunawon 4 · lost 4
GPT-6 Solwon 4 · lost 4
Kimi K2.6won 4 · lost 4
GLM-5.1won 4 · lost 4
Grok 4.5won 4 · lost 4
Mimo V2.5 Prowon 4 · lost 4
Qwen3.7 Maxwon 4 · lost 4
Claude Opus 4.8won 4 · lost 4
GPT-5.5won 4 · lost 4
GPT-5.6 Solwon 4 · lost 4
Gemini 3.8 Flashwon 4 · lost 4
GPT OSS 20Bwon 4 · lost 5
GPT-4owon 4 · lost 5
GPT OSS 120Bwon 4 · lost 5
O3 Miniwon 4 · lost 5
O4 Miniwon 4 · lost 5
Gemini 2.5 Flashwon 4 · lost 5
GPT-4.1won 4 · lost 5
Kimi K2won 4 · lost 5
Claude Sonnet 4.5won 4 · lost 5
Claude 3.5 Haikuwon 3 · lost 5
MiniMax M2won 3 · lost 5
MiniMax M2.7won 3 · lost 5
GPT-5.2won 3 · lost 5
Grok 4.7won 3 · lost 5
Qwen3.5 397B A17Bwon 3 · lost 5
Grok 4.6won 3 · lost 5
GPT-5.6 Terrawon 3 · lost 5
Claude Opus 5won 3 · lost 5
Claude Opus 5.5won 3 · lost 5
Apriel 1.5 15B Thinkerwon 2 · lost 4
Grok 3 Mini Betawon 2 · lost 4
GLM-4.5 Airwon 2 · lost 4
Gemini 2.5 Flash Litewon 2 · lost 4
GPT-4.5won 2 · lost 4
Gemini 3 Prowon 2 · lost 4
GPT-4 Turbowon 3 · lost 6
O1 Miniwon 3 · lost 6
GPT-5 Miniwon 3 · lost 6
Claude Opus 4won 3 · lost 6
GPT-5won 3 · lost 6
GPT-5.1won 3 · lost 6
Gemma 2 27B Itwon 2 · lost 6
Inklingwon 2 · lost 6
Muse Spark 1.1won 2 · lost 6
Claude Sonnet 4won 2 · lost 7
O1won 2 · lost 7
O3won 2 · lost 7
Hy3won 1 · lost 7
Claude Sonnet 5won 1 · lost 7
Claude Opus 4.7won 1 · lost 7
Claude 3 Opuswon 1 · lost 8
GPT-5 Nanowon 1 · lost 8
Claude 3.5 Sonnetwon 1 · lost 8

§ 3 · Sources

Where the numbers come from

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

LMArena Text ↗1 measureread 2026-09-26
LMArena Text 1281
HELM Safety ↗5 measuresread 2026-09-26
HELM Safety · HarmBench 95.8%HELM Safety · SimpleSafetyTests 100%HELM Safety · Anthropic Red Team 99.8%HELM Safety · BBQ 90%HELM Safety · XSTest 85.8%
Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26
Enkrypt · Jailbreak risk 2%Enkrypt · Harmful content risk 6.7%Enkrypt · CBRN risk 3.7%Enkrypt · Toxicity risk 0.4%Enkrypt · Bias risk 78.3%Enkrypt · Insecure code risk 20%

Badge

PublicAI Index badge for Claude 3 Sonnet[![PublicAI Index](https://publicai.io/model-index/badge?model=claude-3-sonnet)](https://publicai.io/model-index/m/claude-3-sonnet)