The LLM benchmark aggregator.

Qwen2.5 72B Instruct

Alibaba · 73B · open weights

Strongest in Safe-prompt compliance (#12 of 82), weakest in Human preference (#232 of 354). Above par in 8 of 12 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 4.8 and Claude Sonnet 5 and ahead of DeepSeek V4 Pro and Kimi K2.6.

§ 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.

Safety56.5−4.9#34/3272/3

Safe-prompt compliance58.4−2#12/821/1

Fairness65.4−4.6#21/2902/2

Toxicity avoidance55.8−3.2#84/2661/1

Harm refusal53.9−7.6#88/2902/2

Jailbreak resistance56.5−8.5#93/2661/1

Secure code51.4−14.8#139/2681/1

Professional49.3−13.1#105/1941/2

Legal50.8−15.1#86/1731/1

Medical46.7−19.8#105/1491/1

Human preference47.4−21.4#232/3541/1

§ 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 Qwen2.5 72B Instruct, left for the other.

OLMo 7B Instructwon 8 · lost 0

SmolLM2 1.7B Instructwon 8 · lost 0

Zephyr 7B Betawon 8 · lost 0

Gemma 7B ITwon 8 · lost 0

Nous Hermes 2 Mixtral 8x7B DPOwon 8 · lost 0

Qwen1.5 14B Chatwon 8 · lost 0

Command Rwon 8 · lost 0

Jamba 1.6 Miniwon 9 · lost 0

Jamba 1.6 Largewon 9 · lost 0

Qwen 2.5 Instruct (7B)won 9 · lost 0

DBRX Instructwon 9 · lost 0

GPT-3.5 Turbowon 9 · lost 0

Mistral Small 3 24B Instructwon 9 · lost 0

Mixtral 8x22B Instructwon 11 · lost 0

Jamba 1.5 Largewon 11 · lost 0

Mixtral 8x7B Instructwon 12 · lost 0

Command R+won 11 · lost 1

Llama 3.1 8B Instructwon 10 · lost 1

Jamba 1.5 Miniwon 10 · lost 1

Llama 3.1 70B Instructwon 10 · lost 1

Gemma 2 9B ITwon 9 · lost 1

Magistral Small 1.2won 8 · lost 1

DeepSeek LLM 67B Chatwon 8 · lost 1

Gemma 2B ITwon 7 · lost 1

Mistral 7B Instruct v0.2won 7 · lost 1

Llama 3.2 3B Instructwon 7 · lost 1

Starling LM 7B Betawon 7 · lost 1

Gemma 2 2B ITwon 7 · lost 1

Amazon Nova Prowon 7 · lost 1

Gemini 1.5 Prowon 10 · lost 2

Llama 3 70B Instructwon 7 · lost 2

Gemini 2.0 Flash-Litewon 7 · lost 2

Ling 3.0 Flashwon 6 · lost 2

Phi-3 Mini 128K Instructwon 6 · lost 2

Phi-3 Mini 4K Instructwon 6 · lost 2

Phi-3 Small 8K Instructwon 6 · lost 2

Phi-3 Medium 4K Instructwon 6 · lost 2

Granite 3.1 8B Instructwon 6 · lost 2

Amazon Nova Litewon 6 · lost 2

Gemma 3 4B ITwon 6 · lost 2

Magistral Mediumwon 6 · lost 2

QwQ 32Bwon 6 · lost 2

GPT-4.1 nanowon 9 · lost 3

DeepSeek V3won 9 · lost 3

Gemini 1.5 Flashwon 8 · lost 3

Llama 4 Maverick Instructwon 8 · lost 3

NVIDIA Nemotron 3 Ultra 550B A55Bwon 8 · lost 3

GLM-5.1won 8 · lost 3

Llama 3.3 70B Instructwon 7 · lost 3

DeepSeek V3.1won 7 · lost 3

Gemma 4 31Bwon 7 · lost 3

Gemini 2.5 Flash-Litewon 6 · lost 3

Llama 4 Scout Instructwon 8 · lost 4

Gemini 2.0 Flashwon 8 · lost 4

Llama 3.1 405B Instructwon 7 · lost 4

Command Awon 7 · lost 4

GLM-4.5won 7 · lost 4

DeepSeek V3.2won 7 · lost 4

Grok 4.1 Fastwon 7 · lost 4

DeepSeek V4 Flashwon 7 · lost 4

GPT-6 Lunawon 7 · lost 4

MiMo V2.5 Prowon 7 · lost 4

Llama 3.2 1B Instructwon 5 · lost 3

Llama 3.1 Tulu 3 8Bwon 5 · lost 3

Aya Expanse 8Bwon 5 · lost 3

Llama 3 8B Instructwon 5 · lost 3

Amazon Nova Microwon 5 · lost 3

Aya Expanse 32Bwon 5 · lost 3

GPT-4won 5 · lost 3

Granite 4.0 H Smallwon 5 · lost 3

Llama 3.1 Nemotron 70B Instructwon 5 · lost 3

Gemma 3 12B ITwon 5 · lost 3

Mercury 2won 5 · lost 3

Gemma 3 27B ITwon 5 · lost 3

Qwen3 30B A3B Instructwon 5 · lost 3

Qwen3.5 27Bwon 5 · lost 3

Qwen3 VL 235B A22B Instructwon 5 · lost 3

Qwen3.5 122B A10Bwon 5 · lost 3

Gemma 4 26B A4Bwon 5 · lost 3

Llama 2 7B Chatwon 6 · lost 4

Llama 2 13B Chatwon 6 · lost 4

Llama 2 70B Chatwon 6 · lost 4

Gemma 2 27B ITwon 6 · lost 4

GPT OSS 20Bwon 7 · lost 5

GPT OSS 120Bwon 7 · lost 5

GPT-4.1 miniwon 7 · lost 5

Grok 3won 7 · lost 5

Grok 4won 7 · lost 5

DeepSeek R1won 7 · lost 5

Kimi K2 Thinkingwon 5 · lost 4

Claude 3 Haikuwon 5 · lost 4

Qwen2 72B Instructwon 5 · lost 4

Claude 3 Sonnetwon 5 · lost 4

Gemini 3 Prowon 5 · lost 4

GPT-4o miniwon 6 · lost 5

Grok 4.3won 6 · lost 5

MiMo V2.6 Flashwon 6 · lost 5

Kimi K2.5won 6 · lost 5

GPT-6 Solwon 6 · lost 5

Kimi K2.6won 6 · lost 5

DeepSeek V4 Prowon 6 · lost 5

NVIDIA Nemotron 3 Super 120B A12Bwon 4 · lost 4

Qwen3.5 35B A3Bwon 4 · lost 4

GPT-5 nanowon 6 · lost 6

O3 Miniwon 6 · lost 6

O4 Miniwon 6 · lost 6

Claude Haiku 4.5won 6 · lost 6

Gemini 2.5 Prowon 6 · lost 6

Mistral Large 4won 5 · lost 6

GLM-4.6won 5 · lost 6

MiniMax M3won 5 · lost 6

Grok 4.5won 5 · lost 6

Qwen3.7 Maxwon 5 · lost 6

GLM-5.3 Flashwon 5 · lost 6

GLM-5.2won 5 · lost 6

Gemini 3.5 Flashwon 5 · lost 6

Gemini 3.6 Flashwon 5 · lost 6

Qwen3.8 Maxwon 5 · lost 6

Gemini 3.8 Flashwon 5 · lost 6

Gemini 2.5 Flashwon 5 · lost 7

Kimi K2won 5 · lost 7

Phi 4won 3 · lost 5

Qwen Maxwon 3 · lost 5

Qwen3.5 397B A17Bwon 3 · lost 5

GPT-5.6 Lunawon 4 · lost 7

GPT-5.6 Terrawon 4 · lost 7

MiMo V2.6 Prowon 4 · lost 7

GPT-5.5won 4 · lost 7

GPT-5.6 Solwon 4 · lost 7

Grok 3 Miniwon 3 · lost 6

Claude Sonnet 4.6won 3 · lost 6

Claude Opus 4.6won 3 · lost 6

Claude 3.7 Sonnetwon 4 · lost 8

Claude Sonnet 4won 4 · lost 8

Qwen3 235B A22B Instructwon 4 · lost 8

Claude 3.5 Haikuwon 3 · lost 7

O1 Miniwon 3 · lost 8

Claude 3.5 Sonnetwon 3 · lost 8

GPT-4.1won 3 · lost 8

MiniMax M2.7won 3 · lost 8

GPT-5.4won 3 · lost 8

Hy3won 2 · lost 6

GPT-4owon 3 · lost 9

GPT-5 miniwon 3 · lost 9

O1won 3 · lost 9

Claude Opus 4won 3 · lost 9

O3won 3 · lost 9

Claude Sonnet 4.5won 3 · lost 9

Claude 3 Opuswon 2 · lost 7

Claude Opus 5won 2 · lost 8

GPT-5.2won 2 · lost 9

Inklingwon 2 · lost 9

Grok 4.6won 2 · lost 9

Claude Opus 4.7won 2 · lost 9

GPT-4 Turbowon 2 · lost 10

GPT-5won 2 · lost 10

GPT-5.1won 2 · lost 10

Muse Spark 1.1won 1 · lost 10

Claude Opus 5.5won 1 · lost 10

MiniMax M2won 0 · lost 8

Grok 4.7won 0 · lost 11

Claude Sonnet 5won 0 · lost 11

Claude Opus 4.8won 0 · lost 11

§ 3 · Sources

Where the numbers come from

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

LMArena Text ↗1 measureread 2026-11-11

LMArena Text 1303

Vals.ai ↗2 measuresread 2026-11-11

Vals · LegalBench 79.4%Vals · MedQA 77.39%

HELM Safety ↗5 measuresread 2026-11-11

HELM Safety · HarmBench 72.8%HELM Safety · SimpleSafetyTests 100%HELM Safety · Anthropic Red Team 99.6%HELM Safety · BBQ 95.4%HELM Safety · XSTest 97.9%

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-11-11

Enkrypt · Jailbreak risk 7.4%Enkrypt · Harmful content risk 28.3%Enkrypt · CBRN risk 8.8%Enkrypt · Toxicity risk 2.2%Enkrypt · Bias risk 44.2%Enkrypt · Insecure code risk 30.2%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=qwen2-5-72b-instruct)](https://publicai.io/model-index/m/qwen2-5-72b-instruct)