The LLM benchmark aggregator.

Olmo 2 32B Instruct

Ai2 · 32B

Strongest in Safe-prompt compliance (#47 of 82), weakest in Fairness (#295 of 299). Above par in 2 of 6 scopes. Among the models it meets almost everywhere, it finishes behind Claude Sonnet 4.5 and GPT-5.1 and ahead of C4ai Aya Expanse 8B and Granite 3.1 8B Instruct.

§ 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.

Safety49.8−11.7#184/3361/3

Safe-prompt compliance52.8−7.6#47/821/1

Harm refusal54.3−7.3#83/2991/2

Fairness34.6−35.4#295/2991/2

Human preference42.8−24.7#253/3411/1

§ 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 Olmo 2 32B Instruct, left for the other.

OLMo 2 13B Instructwon 4 · lost 0

OLMo 2 7B Instructwon 4 · lost 0

GPT-3.5 Turbowon 6 · lost 0

DeepSeek LLM 67B Chatwon 5 · lost 1

DBRX Instructwon 5 · lost 1

Command Rwon 5 · lost 1

Smollm2 1.7B Instructwon 4 · lost 1

Zephyr 7B Betawon 4 · lost 1

Gemma 7B Itwon 4 · lost 1

Mistral 7B Instruct V0.2won 4 · lost 1

Nous Hermes 2 Mixtral 8x7b Dpowon 4 · lost 1

Llama 3.2 3B Instructwon 4 · lost 1

Starling Lm 7B Betawon 4 · lost 1

Granite 3.1 2B Instructwon 4 · lost 1

Qwen1.5 14B Chatwon 4 · lost 1

Gemma 2 2B Itwon 4 · lost 1

Llama 3.1 8B Instructwon 4 · lost 1

Jamba 1.5 Miniwon 4 · lost 1

OLMoE 1B-7B Instructwon 3 · lost 1

Mistral Instruct v0.1 (7B)won 3 · lost 1

Mistral Instruct v0.3 (7B)won 3 · lost 1

Mixtral (8x7B)won 3 · lost 1

Llama 3.1 Instruct Turbo (70B)won 3 · lost 1

Marin 8B Instructwon 3 · lost 1

Mixtral Instruct (8x22B)won 3 · lost 1

Mistral Small 3won 3 · lost 1

IBM Granite 4.0 Microwon 3 · lost 1

DeepSeek R1 (hide reasoning)won 3 · lost 1

IBM Granite 3.3 8B Instructwon 3 · lost 1

Llama 3.1 Instruct Turbo (8B)won 3 · lost 1

Gemma 2B Itwon 3 · lost 2

Phi 3 Mini 128k Instructwon 3 · lost 2

Phi 3 Mini 4k Instructwon 3 · lost 2

Granite 3.0 2B Instructwon 3 · lost 2

Phi 3 Small 8k Instructwon 3 · lost 2

Granite 3.0 8B Instructwon 3 · lost 2

Phi 3 Medium 4k Instructwon 3 · lost 2

Granite 3.1 8B Instructwon 3 · lost 2

C4ai Aya Expanse 8Bwon 3 · lost 2

IBM Granite 4.0 Micro (with guardian)won 2 · lost 2

Palmyra Medwon 2 · lost 2

IBM Granite 4.0 Small (with guardian)won 2 · lost 2

IBM Granite 4.0 Smallwon 2 · lost 2

Palmyra X5won 2 · lost 2

Llama 3 8B Instructwon 3 · lost 3

Qwen1.5 72B Chatwon 3 · lost 3

Command R Pluswon 3 · lost 3

Gemini 2.0 Flash Litewon 3 · lost 3

Llama 3.2 1B Instructwon 2 · lost 3

Llama 3.1 Tulu 3 8Bwon 2 · lost 3

Nova Microwon 2 · lost 3

Mistral Small 24B Instructwon 2 · lost 3

Jamba 1.5 Largewon 2 · lost 3

Gemma 3 4B Itwon 2 · lost 3

Magistral Mediumwon 2 · lost 3

QwQ 32Bwon 2 · lost 3

Command Awon 2 · lost 3

Mistral Smallwon 2 · lost 3

NVIDIA Nemotron 3 Super 120B A12Bwon 2 · lost 3

Qwen3 30B A3B Instructwon 2 · lost 3

GLM-4.5won 2 · lost 3

GLM-4.6won 2 · lost 3

DeepSeek V3.2won 2 · lost 3

Grok 4.1 Fastwon 2 · lost 3

DeepSeek V4 Flashwon 2 · lost 3

MiniMax M3won 2 · lost 3

Grok 4.3won 2 · lost 3

Kimi K2.5won 2 · lost 3

Gemma 4 31Bwon 2 · lost 3

Mimo V2.6 Flashwon 2 · lost 3

Kimi K2.6won 2 · lost 3

Mimo V2.5 Prowon 2 · lost 3

Qwen3.7 Maxwon 2 · lost 3

GLM-5.2won 2 · lost 3

Gemini 3.5 Flashwon 2 · lost 3

Gemini 3.6 Flashwon 2 · lost 3

Claude 3 Haikuwon 2 · lost 4

Gemini 1.5 Flashwon 2 · lost 4

Llama 4 Scout Instructwon 2 · lost 4

Gemini 1.5 Prowon 2 · lost 4

Gemini 2.0 Flashwon 2 · lost 4

DeepSeek V3won 2 · lost 4

Grok 3won 2 · lost 4

Grok 4won 2 · lost 4

DeepSeek R1won 2 · lost 4

Llama 3.1 Instruct Turbo (405B)won 1 · lost 3

Qwen 2.5 Instruct Turbo (7B)won 1 · lost 3

Gemini 2.0 Prowon 1 · lost 3

Qwen3 235B A22B Throughputwon 1 · lost 3

Palmyra Xwon 1 · lost 3

Qwen 2.5 Instruct Turbo (72B)won 1 · lost 3

Gemma 2 9B Itwon 1 · lost 4

C4ai Aya Expanse 32Bwon 1 · lost 4

Gemma 2 27B Itwon 1 · lost 4

Qwen2.5 72B Instructwon 1 · lost 4

Mistral Largewon 1 · lost 4

Qwen Maxwon 1 · lost 4

Gemma 3 12B Itwon 1 · lost 4

Gemma 3 27B Itwon 1 · lost 4

Qwen3.5 35B A3Bwon 1 · lost 4

Qwen3.5 27Bwon 1 · lost 4

Qwen3 Vl 235B A22B Instructwon 1 · lost 4

Qwen3.5 122B A10Bwon 1 · lost 4

Gemma 4 26B A4Bwon 1 · lost 4

GPT-6 Lunawon 1 · lost 4

Qwen3.5 397B A17Bwon 1 · lost 4

Grok 4.6won 1 · lost 4

GPT-5.6 Lunawon 1 · lost 4

GPT-6 Solwon 1 · lost 4

DeepSeek V4 Prowon 1 · lost 4

GLM-5.1won 1 · lost 4

Grok 4.5won 1 · lost 4

GLM-5.3 Flashwon 1 · lost 4

Qwen3.8 Maxwon 1 · lost 4

Mimo V2.6 Prowon 1 · lost 4

Claude Opus 5won 1 · lost 4

Gemini 3.8 Flashwon 1 · lost 4

Llama 3 70B Instructwon 1 · lost 5

Claude 3 Sonnetwon 1 · lost 5

GPT-4o Miniwon 1 · lost 5

GPT OSS 20Bwon 1 · lost 5

Claude 3 Opuswon 1 · lost 5

GPT-4.1 Nanowon 1 · lost 5

Llama 4 Maverick Instructwon 1 · lost 5

GPT OSS 120Bwon 1 · lost 5

O3 Miniwon 1 · lost 5

Grok 3 Miniwon 1 · lost 5

GLM-4.5 Airwon 1 · lost 5

Gemini 2.5 Flash Litewon 1 · lost 5

GPT-4.1 Miniwon 1 · lost 5

Claude Sonnet 4won 1 · lost 5

Gemini 2.5 Flashwon 1 · lost 5

Claude Haiku 4.5won 1 · lost 5

GPT-4.5won 1 · lost 5

Gemini 2.5 Prowon 1 · lost 5

Gemini 3 Prowon 1 · lost 5

Palmyra Finwon 0 · lost 4

Phi 4won 0 · lost 5

Nova Litewon 0 · lost 5

GPT-4won 0 · lost 5

Nova Prowon 0 · lost 5

Claude 3.5 Haikuwon 0 · lost 5

MiniMax M2won 0 · lost 5

MiniMax M2.7won 0 · lost 5

GPT-5.2won 0 · lost 5

Grok 4.7won 0 · lost 5

Inklingwon 0 · lost 5

Hy3won 0 · lost 5

Claude Sonnet 5won 0 · lost 5

GPT-5.6 Terrawon 0 · lost 5

Claude Sonnet 4.6won 0 · lost 5

GPT-5.4won 0 · lost 5

Claude Opus 4.8won 0 · lost 5

GPT-5.5won 0 · lost 5

GPT-5.6 Solwon 0 · lost 5

Muse Spark 1.1won 0 · lost 5

Claude Opus 4.7won 0 · lost 5

Claude Opus 4.6won 0 · lost 5

Claude Opus 5.5won 0 · lost 5

Qwen2 72B Instructwon 0 · lost 6

GPT-4 Turbowon 0 · lost 6

GPT-4owon 0 · lost 6

O1 Miniwon 0 · lost 6

GPT-5 Nanowon 0 · lost 6

Qwen3 Next 80B A3B Thinkingwon 0 · lost 6

Claude 3.5 Sonnetwon 0 · lost 6

Claude 3.7 Sonnetwon 0 · lost 6

GPT-5 Miniwon 0 · lost 6

O4 Miniwon 0 · lost 6

O1won 0 · lost 6

GPT-4.1won 0 · lost 6

Kimi K2won 0 · lost 6

Qwen3 235B A22B Instructwon 0 · lost 6

Claude Opus 4won 0 · lost 6

O3won 0 · lost 6

GPT-5won 0 · lost 6

GPT-5.1won 0 · lost 6

Claude Sonnet 4.5won 0 · lost 6

§ 3 · Sources

Where the numbers come from

2 publications, 6 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-09-27

LMArena Text 1252

HELM Safety ↗5 measuresread 2026-09-27

HELM Safety · HarmBench 84.1%HELM Safety · SimpleSafetyTests 98%HELM Safety · Anthropic Red Team 99.3%HELM Safety · BBQ 71.4%HELM Safety · XSTest 95.4%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=olmo-2-32b-instruct)](https://publicai.io/model-index/m/olmo-2-32b-instruct)