Magistral Small 1.2

Name: Magistral Small 1.2
Author: Mistral AI

Mistral AI · Providers EU

Catalogue only Open weights European provider

Sovereignty and hosting

Catalogue only

This model is catalogued for reference. It is not hosted on LLM EU yet, and no first-party endpoint serves it here. Requests to it go to the provider directly, under that provider's own terms.

Summary

Magistral Small 1.2 is a 24B Apache-2.0 reasoning model derived from Mistral Small 3.2, with a vision encoder and explicit think tokens for the reasoning trace. The vendor reports AIME24 at 86.14%, AIME25 at 77.34% and GPQA Diamond at 70.07%. Mistral notes that quality may degrade beyond roughly 40k tokens even though the window is 128k.

Specifications

Provider: Mistral AI (FR)
Context window: 131k
Parameters: 24.01B
Modalities: text, vision
License: apache-2.0 · licence
Open weights: yes
Release date: 12 Sept 2025
Tasks: reasoning, coding, summarization

Languages

Strong: en

Adequate: fr, de, es, it, pt, nl, pl, ru, zh, ja, ko, ar

Languages

Strong

Adequate

Not verified: the provider has not documented this claim. — a language is listed as strong only where a source claims it; the card links the evidence where one exists.

When to use it

Strengths

visible reasoning traces through dedicated think tokens
maths and science scores competitive at the 24B size class
vision input inherited from Mistral Small 3.2
Apache-2.0 licence

Limits

vendor states quality may degrade beyond about 40k tokens
all benchmark figures are self-reported
no evaluation has been run by LLM EU

Pricing

This model is catalogued for reference. It is not hosted on LLM EU yet, and no first-party endpoint serves it here. Requests to it go to the provider directly, under that provider's own terms.

Verification status

Where a field has no source, the card shows it as unverified instead of filling it in. An empty field is information.

LLM EU Verified means documented, not legally certified. It is not an AI Act conformity assessment and LLM EU is not an EU body.

Versions and variants

catalogue · Catalogue only · weights

Example request

curl https://api.llmeu.com/v1/chat/completions \
  -H "Authorization: Bearer $LLMEU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"magistral-small-1-2","messages":[{"role":"user","content":"Hello"}]}'

No playground is deployed for this model: no endpoint serves it, so a request would have nowhere to go.

Benchmarks

AIME25 pass@1: 77.34% (vendor-reported) source
GPQA Diamond: 70.07% (vendor-reported) source
LiveCodeBench v5: 70.88% (vendor-reported) source

Sources

huggingface.co

Card changelog

12 Sept 2025 — Release date
19 Sept 2026 — Last reviewed

Last reviewed: 19 Sept 2026

Tasks

Reasoning