ministral-3:3b

Upstream reference: Ministral 3 3B Instruct 2512

Publisher: Mistral AI

Mistral describes this instruction-tuned text-and-image model as the smallest Ministral 3 release, intended for local and edge applications.

Publisher-described context, separate from Sigma5C observations and recorded catalog statistics.

Published upstream specifications

Language component

3.4 billion parameters Source

Vision encoder

0.4 billion parameters Source

Published checkpoint precision

FP8 Source

Published context

256k tokens, using publisher notation Source

Upstream license label

Apache-2.0 Source

Publisher-described uses

• Image captioning, text classification, translation, data extraction and short-form text generation named by Mistral. Source

Reference limitations

A recorded tag and an upstream card association do not verify the deployed checkpoint, quantization, configured context or current hosted availability.

Publisher sources

• Ministral 3 3B Instruct 2512 model card — Mistral AI. Checked 2026-11-02.

Links

Recorded references do not establish downloadable weights, licensing or current availability.

• Mistral AI — official site
• Ollama
• Hugging Face
• Official model announcement / documentation

Where ministral-3:3b comes from

Named in 2 observation pages

• 3B-7B local models — all 53 · V1
• Models as data workers · V1