The LLM benchmark aggregator.

Claude 3 Haiku

Anthropic

Strongest in Harm refusal (#21 of 300), weakest in Fairness (#297 of 300). Above par in 5 of 13 scopes. Among the models it meets almost everywhere, it finishes behind Claude Sonnet 5 and Claude 3 Opus and ahead of Mimo V2.6 Flash and Gemini 2.0 Flash Lite.

§ 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.

Safety51.3−10.2#149/3372/3

Harm refusal58.4−3.2#21/3002/2

Toxicity avoidance58.1−0.9#26/2721/1

Jailbreak resistance60.3−4.6#41/2721/1

Safe-prompt compliance30.4−30#77/821/1

Secure code55.3−11.2#114/2741/1

Fairness34.4−35.6#297/3002/2

Coding40.3−28.8#152/1651/5

Code generation31.3−38.7#74/771/2

Core abilities43.7−23.7#165/2041/3

General intelligence40.9−29.2#168/2041/3

Human preference43.7−23.8#250/3421/1

§ 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 3 Haiku, left for the other.

DBRX Instructwon 8 · lost 1

GPT-3.5 Turbowon 8 · lost 1

Magistral Smallwon 7 · lost 1

Jamba Mini 1.7won 7 · lost 1

Jamba Large 1.7won 7 · lost 1

Gemma 2B Itwon 7 · lost 1

Smollm2 1.7B Instructwon 7 · lost 1

Zephyr 7B Betawon 7 · lost 1

Gemma 7B Itwon 7 · lost 1

Mistral 7B Instruct V0.2won 7 · lost 1

Nous Hermes 2 Mixtral 8x7b Dpowon 7 · lost 1

Llama 3.2 3B Instructwon 7 · lost 1

Starling Lm 7B Betawon 7 · lost 1

Qwen1.5 14B Chatwon 7 · lost 1

Phi 3 Medium 4k Instructwon 7 · lost 1

Llama 3.1 8B Instructwon 7 · lost 1

Grok Build 0.1won 8 · lost 2

Llama 3 8B Instructwon 7 · lost 2

Llama 3.2 1B Instructwon 6 · lost 2

Granite 3.0 2B Instructwon 6 · lost 2

Granite 3.1 2B Instructwon 6 · lost 2

Granite 3.0 8B Instructwon 6 · lost 2

Gemma 2 2B Itwon 6 · lost 2

Granite 3.1 8B Instructwon 6 · lost 2

Llama 3.1 Tulu 3 8Bwon 6 · lost 2

Jamba 1.5 Miniwon 6 · lost 2

Nova Microwon 6 · lost 2

Nova Litewon 6 · lost 2

Command Awon 8 · lost 3

Nemotron 3 Ultra 550B A55Bwon 7 · lost 3

Gemma 3 4B Itwon 7 · lost 3

Magistral Mediumwon 7 · lost 3

DeepSeek LLM 67B Chatwon 6 · lost 3

Grok 4.1 Fastwon 6 · lost 3

Ling 3.0 Flash VLwon 5 · lost 3

Phi 3 Mini 128k Instructwon 5 · lost 3

Phi 3 Mini 4k Instructwon 5 · lost 3

Phi 3 Small 8k Instructwon 5 · lost 3

Gemma 2 9B Itwon 5 · lost 3

Mistral Small 24B Instructwon 5 · lost 3

Jamba 1.5 Largewon 5 · lost 3

Qwen3 30B A3B Instructwon 5 · lost 3

Grok 4.3won 7 · lost 5

Command R Pluswon 5 · lost 4

QwQ 32Bwon 5 · lost 4

Gemini 1.5 Prowon 5 · lost 4

Gemini 2.0 Flash Litewon 5 · lost 4

Mimo V2.6 Flashwon 5 · lost 4

C4ai Aya Expanse 8Bwon 4 · lost 4

Nova Prowon 4 · lost 4

Mistral Largewon 4 · lost 4

Gemma 3 12B Itwon 4 · lost 4

NVIDIA Nemotron 3 Super 120B A12Bwon 4 · lost 4

Gemma 4 26B A4Bwon 4 · lost 4

Kimi K2.7 Codewon 5 · lost 5

Mistral Smallwon 5 · lost 5

GLM-4.5won 5 · lost 5

Grok 3won 5 · lost 5

GPT-4o Miniwon 6 · lost 7

Llama 4 Scout Instructwon 5 · lost 6

Gemma 4 12Bwon 4 · lost 5

Gemini 1.5 Flashwon 4 · lost 5

GPT-4.1 Nanowon 5 · lost 7

Grok 4won 5 · lost 7

DeepSeek V4 Flashwon 5 · lost 7

MiniMax M3won 5 · lost 7

Kimi K2.6won 5 · lost 7

DeepSeek V4 Prowon 5 · lost 7

GLM-5.3 Flashwon 5 · lost 7

Gemini 3.6 Flashwon 5 · lost 7

DeepSeek V3won 5 · lost 8

DeepSeek R1won 5 · lost 8

Gemini 2.5 Prowon 5 · lost 8

Phi 4won 3 · lost 5

C4ai Aya Expanse 32Bwon 3 · lost 5

GPT-4won 3 · lost 5

Qwen3.5 35B A3Bwon 3 · lost 5

Qwen3.5 27Bwon 3 · lost 5

Qwen3 Vl 235B A22B Instructwon 3 · lost 5

GPT OSS 20Bwon 4 · lost 7

Gemma 3 27B Itwon 4 · lost 7

GLM-4.6won 4 · lost 7

DeepSeek V3.2won 4 · lost 7

Kimi K2.5won 4 · lost 7

Mimo V2.5 Prowon 4 · lost 7

Mimo V2.6 Prowon 4 · lost 7

Kimi K2 Thinkingwon 3 · lost 6

Qwen Maxwon 3 · lost 6

GLM-5.1won 3 · lost 6

Llama 4 Maverick Instructwon 4 · lost 8

Gemini 2.0 Flashwon 4 · lost 8

Gemma 4 31Bwon 4 · lost 8

Claude Sonnet 4.5won 4 · lost 8

Qwen3.7 Maxwon 4 · lost 8

GLM-5.2won 4 · lost 8

Gemini 3.5 Flashwon 4 · lost 8

Gemini 2.5 Flashwon 4 · lost 9

MiniMax M2won 3 · lost 7

Qwen3.5 122B A10Bwon 3 · lost 7

Qwen3.5 397B A17Bwon 3 · lost 7

Llama 3 70B Instructwon 3 · lost 8

GPT-4owon 3 · lost 8

Gemma 2 27B Itwon 2 · lost 6

Qwen2.5 72B Instructwon 2 · lost 6

GPT OSS 120Bwon 3 · lost 9

GPT-4.1 Miniwon 3 · lost 9

Claude Haiku 4.5won 3 · lost 9

GPT-4.1won 3 · lost 9

Kimi K2won 3 · lost 9

GPT-6 Lunawon 3 · lost 9

Grok 4.6won 3 · lost 9

GPT-5.6 Lunawon 3 · lost 9

GPT-6 Solwon 3 · lost 9

Grok 4.5won 3 · lost 9

GPT-5.6 Terrawon 3 · lost 9

GPT-5.4won 3 · lost 9

Qwen3.8 Maxwon 3 · lost 9

GPT-5.5won 3 · lost 9

GPT-5.6 Solwon 3 · lost 9

Claude Opus 5won 3 · lost 9

Gemini 3.8 Flashwon 3 · lost 9

Claude Opus 5.5won 3 · lost 9

O4 Miniwon 3 · lost 10

Qwen3 235B A22B Instructwon 3 · lost 10

Claude Opus 4won 3 · lost 10

Qwen2 72B Instructwon 2 · lost 7

Claude 3.5 Haikuwon 2 · lost 7

MiniMax M2.7won 2 · lost 7

Gemini 3 Prowon 2 · lost 7

O1 Miniwon 2 · lost 8

Claude 3.7 Sonnetwon 2 · lost 8

GPT-5.1won 2 · lost 8

Claude Sonnet 4.6won 2 · lost 8

GPT-4 Turbowon 2 · lost 9

O3 Miniwon 2 · lost 9

GPT-5.2won 2 · lost 10

Grok 4.7won 2 · lost 10

Claude Opus 4.8won 2 · lost 10

O3won 2 · lost 11

Qwen3 Next 80B A3B Thinkingwon 1 · lost 7

GLM-4.5 Airwon 1 · lost 7

Gemini 2.5 Flash Litewon 1 · lost 7

O1won 1 · lost 9

Hy3won 1 · lost 9

Claude Opus 4.6won 1 · lost 9

GPT-5 Nanowon 1 · lost 10

Claude 3.5 Sonnetwon 1 · lost 10

GPT-5 Miniwon 1 · lost 11

GPT-5won 1 · lost 11

Inklingwon 1 · lost 11

Muse Spark 1.1won 1 · lost 11

Claude Opus 4.7won 1 · lost 11

Claude Sonnet 4won 1 · lost 12

Claude 3 Sonnetwon 0 · lost 9

Claude 3 Opuswon 0 · lost 9

Claude Sonnet 5won 0 · lost 12

§ 3 · Sources

Where the numbers come from

5 publications, 14 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-09-26

LMArena Text 1262

LiveCodeBench ↗1 measureread 2026-09-26

LiveCodeBench 20.2%

Kagi LLM Benchmark ↗1 measureread 2026-09-26

Kagi LLM Benchmark 34.2%

HELM Safety ↗5 measuresread 2026-09-26

HELM Safety · HarmBench 91.3%HELM Safety · SimpleSafetyTests 100%HELM Safety · Anthropic Red Team 100%HELM Safety · BBQ 62.5%HELM Safety · XSTest 85.3%

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26

Enkrypt · Jailbreak risk 4.1%Enkrypt · Harmful content risk 12.8%Enkrypt · CBRN risk 7.3%Enkrypt · Toxicity risk 0.6%Enkrypt · Bias risk 87.1%Enkrypt · Insecure code risk 22.7%

Badge

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