The LLM benchmark aggregator.

Magistral Small 1.2

Mistral

Strongest in Mathematics (#66 of 152, on 1 of its 2 boards), weakest in Safety (#290 of 329). Above par in 2 of 17 scopes. Among the models it meets almost everywhere, it finishes behind Claude Opus 5 and GPT-5.6 Sol and ahead of GPT-4.1 nano and Mistral Medium 3.1.

§ 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.

Reasoning47.7−17.7#117/1951/5

Mathematics53.8−10#66/1521/2

Science37.6−25.9#102/1221/1

Professional42.4−20#171/1941/2

Medical49.4−17.1#91/1491/1

Finance41.1−21.3#135/1621/1

Legal35−30.9#172/1731/1

Knowledge40.1−22.4#180/1971/3

Academic knowledge26.1−37.8#120/1231/1

Agents43.2−23.2#244/2981/6

Knowledge work36.4−34.3#171/2081/2

Safety44−17.4#290/3291/3

Fairness47.4−22.6#149/2921/2

Jailbreak resistance50.6−14.4#174/2681/1

Secure code35.3−30.9#231/2701/1

Toxicity avoidance40.8−18.2#238/2681/1

Harm refusal43.7−17.8#258/2921/2

§ 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 Magistral Small 1.2, left for the other.

Mistral Small 3.1 24B Instructwon 7 · lost 4

Mistral Medium 3.5won 7 · lost 4

Command R+won 8 · lost 5

Jamba 1.6 Miniwon 9 · lost 6

Jamba 1.6 Largewon 9 · lost 6

Llama 4 Scout Instructwon 10 · lost 7

Llama 3.1 8B Instructwon 6 · lost 5

Mistral Medium 3.1won 6 · lost 5

GPT-4.1 nanowon 9 · lost 8

GPT-4o miniwon 8 · lost 8

Gemini 1.5 Prowon 7 · lost 8

DeepSeek R1won 7 · lost 9

Llama 3.3 70B Instructwon 7 · lost 9

Command Awon 7 · lost 9

Gemini 1.5 Flashwon 5 · lost 9

DeepSeek V3won 6 · lost 11

Llama 4 Maverick Instructwon 6 · lost 11

Mistral Large 3won 4 · lost 8

Grok 3won 5 · lost 10

Grok 4.1 Fastwon 5 · lost 11

Gemma 4 31Bwon 4 · lost 9

GPT-4owon 5 · lost 12

Qwen3.8 27Bwon 3 · lost 8

Gemini 3.5 Flash-Litewon 3 · lost 8

MiMo V2.5won 3 · lost 8

Gemini 2.0 Flashwon 4 · lost 11

Claude 3.5 Haikuwon 4 · lost 12

Kimi K2won 4 · lost 12

Grok 4.3won 4 · lost 13

NVIDIA Nemotron 3 Ultra 550B A55Bwon 4 · lost 13

MiMo V2.5 Prowon 4 · lost 13

Gemini 2.5 Flash-Litewon 3 · lost 10

Mistral Large 4won 3 · lost 11

MiMo V2.6 Flashwon 3 · lost 11

Ling 3.0 Flashwon 3 · lost 13

Gemini 2.5 Flashwon 3 · lost 13

Grok 4won 3 · lost 13

Qwen3.6 Pluswon 2 · lost 9

Inkling Smallwon 2 · lost 9

O3 Miniwon 3 · lost 14

Qwen3.7 Maxwon 3 · lost 14

GLM-5.3 Flashwon 3 · lost 14

Gemini 3.6 Flashwon 3 · lost 14

Kimi K2.6won 3 · lost 14

GLM-5.2won 3 · lost 14

DeepSeek V4 Flashwon 3 · lost 14

MiniMax M3won 3 · lost 14

Gemini 3.1 Flash-Litewon 2 · lost 10

Claude 3.5 Sonnetwon 2 · lost 12

MiMo V2.6 Prowon 2 · lost 12

Claude Opus 4won 2 · lost 13

GPT-6 Lunawon 2 · lost 13

Claude 3.7 Sonnetwon 2 · lost 14

GPT-5 nanowon 2 · lost 14

Qwen3 235B A22B Instructwon 2 · lost 14

GPT-4.1won 2 · lost 14

Inklingwon 2 · lost 15

Gemini 2.5 Prowon 2 · lost 15

GPT OSS 20Bwon 2 · lost 15

Claude Sonnet 4won 2 · lost 15

Gemini 3.5 Flashwon 2 · lost 15

GPT-4.1 miniwon 2 · lost 15

Gemini 3.8 Flashwon 2 · lost 15

Kimi K2.5won 2 · lost 15

DeepSeek V4 Prowon 2 · lost 15

GLM-4.6won 2 · lost 15

Magistral Medium 1.2won 1 · lost 10

GPT-5.4 miniwon 1 · lost 10

Gemini 3.7 Flashwon 1 · lost 10

Kimi K3won 1 · lost 10

GLM-5.3won 1 · lost 10

Gemini 3.1 Prowon 1 · lost 11

GPT-5.4 nanowon 1 · lost 11

GLM-4.7won 1 · lost 11

Grok 4.7won 1 · lost 13

O1won 1 · lost 14

GLM-4.5won 1 · lost 14

Claude Sonnet 4.6won 1 · lost 14

O4 Miniwon 1 · lost 15

GPT OSS 120Bwon 1 · lost 16

MiniMax M2.7won 1 · lost 16

Grok 4.5won 1 · lost 16

GLM-5.1won 1 · lost 16

Qwen3.8 Maxwon 1 · lost 16

Claude Haiku 4.5won 1 · lost 16

DeepSeek V3.2won 1 · lost 16

GPT-5.4won 1 · lost 16

Grok 4.6won 1 · lost 16

Muse Spark 1.1won 1 · lost 16

GPT-5.6 Lunawon 1 · lost 16

Muse Sparkwon 0 · lost 11

Claude Opus 4.5won 0 · lost 11

Claude Fable 5won 0 · lost 11

Claude Fable 5.1won 0 · lost 11

Grok 3 Miniwon 0 · lost 12

Claude Opus 4.6won 0 · lost 13

Gemini 3 Prowon 0 · lost 13

Kimi K2 Thinkingwon 0 · lost 15

GPT-6 Solwon 0 · lost 15

Claude Opus 5.5won 0 · lost 15

O3won 0 · lost 16

GPT-5.2won 0 · lost 16

GPT-5 miniwon 0 · lost 17

GPT-5.1won 0 · lost 17

GPT-5won 0 · lost 17

GPT-5.6 Terrawon 0 · lost 17

GPT-5.5won 0 · lost 17

Claude Opus 4.7won 0 · lost 17

Claude Sonnet 4.5won 0 · lost 17

Claude Opus 4.8won 0 · lost 17

Claude Sonnet 5won 0 · lost 17

GPT-5.6 Solwon 0 · lost 17

Claude Opus 5won 0 · lost 17

§ 3 · Sources

Where the numbers come from

3 publications, 15 figures. Every one links to the page it was read from.

GDPval-AA ↗1 measureread 2026-11-10

GDPval-AA 12

Vals.ai ↗8 measuresread 2026-11-10

Vals · LegalBench 40.02%Vals · CorpFin 44.02%Vals · TaxEval 60.3%Vals · MortgageTax 62.12%Vals · MedQA 82.36%Vals · GPQA Diamond 58.33%Vals · MMLU Pro 62.13%Vals · AIME 80.68%

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-11-10

Enkrypt · Jailbreak risk 12.3%Enkrypt · Harmful content risk 63.9%Enkrypt · CBRN risk 15.5%Enkrypt · Toxicity risk 12.5%Enkrypt · Bias risk 81.9%Enkrypt · Insecure code risk 63.1%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=magistral-small-1-2)](https://publicai.io/model-index/m/magistral-small-1-2)