Mistral
6 MLX aliases · Mistral's general + code lines — Ministral / Mistral / Mistral-Small-4 / Devstral.
Pick one
One command per line of this family, smallest download first. Your Mac needs the download size in free memory plus room for macOS and the context window; the hardware tiers page has the engine's picks for every RAM size. The first run downloads the weights and starts an OpenAI-compatible server on http://localhost:8000/v1.
rapid-mlx serve devstral-24b-4bitrapid-mlx serve mistral-24b-4bit
The Mistral line on rapid-mlx covers Mistral's general-purpose 3B and 24B chat models, the 119B Mistral-Small-4, and the Devstral code-tuned 24B variants (V1 + V2). All use the hermes tool envelope so OpenAI-compatible clients work without per-model config.
- family
- Mistral
- aliases
- 6
- lines
- 2
- install
- rapid-mlx serve <alias>
- OpenAI base URL
- http://localhost:8000/v1
Download
Every alias on this page downloads with one command — the pull buttons in the tables below copy it. 3 of the 6 aliases on this page are mirrored on the rapid-mlx CDN; the rest pull from Hugging Face directly — with automatic mid-pull fallback to Hugging Face if a mirror file slows down. Weights land in the standard Hugging Face cache, and rapid-mlx serve pulls automatically on first use. Live mirror status →
Lines in this family
Mistral (chat) · 4 aliases
Ministral 3B and Mistral 24B — general-purpose chat models.
parser: hermes
| alias | hf repo | tool parser | reasoning | flags | context | AA index | get it |
|---|---|---|---|---|---|---|---|
| mistral-24b-4bit | mlx-community/Mistral-Small-3.1-24B-Instruct-2503-4bit | mistral | — | spec | 128K | 14.9 | CDN |
| mistral-small-4-119b | mlx-community/Mistral-Small-4-119B-2603-4bit | mistral | — | spec | — | 19.7 | HF |
| mistral-small-4-119b-4bit | mlx-community/Mistral-Small-4-119B-2603-4bit | mistral | — | spec | — | 19.7 | HF |
| mistral-small-4-119b-8bit | mlx-community/Mistral-Small-4-119B-2603-8bit | mistral | — | spec | — | 19.7 | HF |
Devstral · 2 aliases
Code-tuned Mistral 24B in two generations — V1 and V2. Devstral V2 is strictly better than V1 on SWE-Bench; we keep V1 in the registry for reproducibility.
parser: hermes
| alias | hf repo | tool parser | reasoning | flags | context | AA index | get it |
|---|---|---|---|---|---|---|---|
| devstral-24b-4bit | mlx-community/Devstral-Small-2507-4bit | mistral | — | spec | 128K | 9.1 | CDN |
| devstral-v2-24b-4bit | mlx-community/Devstral-Small-2-24B-Instruct-2512-4bit | mistral | — | spec | 384K | 17.7 | CDN |
Notes & caveats
- Devstral V2 is the recommended pick for code on a 32 GB+ Mac — strictly better than V1 on SWE-Bench.
Context is read from the config.json of the exact build each alias pulls; for an embedding model it is the most input tokens the engine embeds, which can be less than the config declares. AA index is the Artificial Analysis Intelligence Index for the base model at full precision with reasoning on — a property of the model, not a score for our quantised build.