Mistral Small 4
Language modelOpen weightsCard summarizedOur powerful hybrid model unifying instruct, reasoning, and coding capabilities in a single model
Active● Re-checked 12 Oct 2026 · 27 facts · 0 corrections · 1 unconfirmedSource: Introducing Mistral Small 4 and Hugging Face model card
Use it via API
Where to call it
- Mistral API
mistral-small-2603 - Self-host (weights)
mistralai/Mistral-Small-4-119B-2603
Good for
- Coding automation, codebase exploration and coding agents
- General chat assistants, document understanding and multimodal analysis
- Math, research and complex reasoning
Overview
Small 4
Our powerful hybrid model unifying instruct, reasoning, and coding capabilities in a single model
- Cost
- $0.15 in
- Context
- 256K
- Input / output
- $0.15 / $0.60
- Max output
- —
- Knows up to
- —
- Released
- 16 Mar 2026
The first Mistral model to unify the capabilities of its Magistral (reasoning), Pixtral (vision) and Devstral (coding) models. A 119B mixture-of-experts model with about 6B active parameters, 256K context, text and image input, and per-request reasoning. Apache 2.0 open weights. launch post
mistral-small-2603- Input / output per 1M
- $0.15 / $0.60
- Context window
- 256K
- Max output
- —
- Knows up to
- —
- Reasoning
- Configurable (reasoning_effort: none or high)
- Weights
- Open · Apache 2.0
About this card. A model card is the maker's own account of a model: what it's for, how it was trained and tested, and where it falls short. This page condenses Mistral AI's Introducing Mistral Small 4 and Hugging Face model card. Every claim carries its source page.