IBM · First MLX build 31 Mar 2026
granite-vision-3.2-2b
MLX
TextVision apache-2.0
A 3.1B parameter vision and text model from IBM, with 8 MLX builds, from 4-bit to bf16.
- Parameters
- 3.1B From Hugging Face
- Quantisation
- 4-bit granite-vision-3.2-2b-4bit
- Memory needed
- 5 GB Estimate, unified memory
- Download
- 2.3 GB Weights, estimated
- Downloads
- 78 This build, last 30 days
- Task
- Vision and text Text, Vision
- MLX builds
- 8 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 5 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/granite-vision-3.2-2b-4bit --image photo.jpg --prompt "Describe this image"MLX builds
8 on mlx-community| granite-vision-3.2-2b-4bit | 4-bit | 2.3 GB | 5 GB | 78 |
|---|---|---|---|---|
| granite-vision-3.2-2b-nvfp4 | 4-bit | 2.3 GB | 5 GB | 69 |
| granite-vision-3.2-2b-mxfp4 | MXFP4 | 2.3 GB | 5 GB | 47 |
| granite-vision-3.2-2b-5bit | 5-bit | 2.7 GB | 5 GB | 35 |
| granite-vision-3.2-2b-6bit | 6-bit | 3.0 GB | 6 GB | 38 |
| granite-vision-3.2-2b-8bit | 8-bit | 3.7 GB | 7 GB | 43 |
| granite-vision-3.2-2b-mxfp8 | MXFP8 | 3.7 GB | 7 GB | 35 |
| granite-vision-3.2-2b-bf16 | bf16 | 6.2 GB | 9 GB | 45 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.