Mistral AI · First MLX build 9 Dec 2025
Devstral-Small-2-24B-Instruct-2512
MLX
TextVision apache-2.0 Trending
A 24B parameter vision and text model from Mistral AI, with 12 MLX builds, from 4-bit to bf16.
- Parameters
- 24B From Hugging Face
- Quantisation
- 4-bit Devstral-Small-2-24B-Instruct-2512-4bit
- Memory needed
- 19 GB Estimate, unified memory
- Download
- 15.1 GB Weights, estimated
- Downloads
- 72K This build, last 30 days
- Task
- Vision and text Text, Vision
- MLX builds
- 12 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 19 GB at 4-bit
8 GB
16 GB
24 GB
32 GB
64 GB
128 GB
Fits with headroomTight, close other appsWon’t fit
Estimated from the size of the weights, plus a fifth for context and working memory and 2 GB for macOS and your apps. Long contexts need more.
Run it
Terminal · mlx-vlm$ pip install -U mlx-vlm
$ python -m mlx_vlm.generate --model mlx-community/Devstral-Small-2-24B-Instruct-2512-4bit --image photo.jpg --prompt "Describe this image"MLX builds
12 on mlx-community| Devstral-Small-2-24B-Instruct-2512-OptiQ-4bit | 4-bit | 14.1 GB | 18 GB | 1.2K |
|---|---|---|---|---|
| Devstral-Small-2-24B-Instruct-2512-4bit | 4-bit | 15.1 GB | 19 GB | 72K |
| mistralai_Devstral-Small-2-24B-Instruct-2512-MLX-4Bit | 4-bit | 15.1 GB | 19 GB | 1.0K |
| mistralai_Devstral-Small-2-24B-Instruct-2512-MLX-MXFP4 | MXFP4 | 15.1 GB | 19 GB | 935 |
| mistralai_Devstral-Small-2-24B-Instruct-2512-MLX-5Bit | 5-bit | 18.0 GB | 23 GB | 806 |
| Devstral-Small-2-24B-Instruct-2512-5bit | 5-bit | 18.0 GB | 23 GB | 175 |
| Devstral-Small-2-24B-Instruct-2512-6bit | 6-bit | 20.8 GB | 26 GB | 971 |
| mistralai_Devstral-Small-2-24B-Instruct-2512-MLX-6Bit | 6-bit | 20.8 GB | 26 GB | 777 |
| mistralai_Devstral-Small-2-24B-Instruct-2512-MLX-8Bit | 8-bit | 26.6 GB | 32 GB | 1.4K |
| Devstral-Small-2-24B-Instruct-2512-8bit | 8-bit | 26.6 GB | 32 GB | 1.2K |
| mistralai_Devstral-Small-2-24B-Instruct-2512-MLX-BF16 | bf16 | 48.0 GB | 56 GB | 649 |
| Devstral-Small-2-24B-Instruct-2512-bf16 | bf16 | 48.0 GB | 56 GB | 296 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.