IBM · First MLX build 1 Apr 2026
granite-4.0-3b-vision
MLX
TextVision apache-2.0
A 4.3B parameter vision and text model from IBM, with 7 MLX builds, from 4-bit to MXFP8.
- Parameters
- 4.3B From Hugging Face
- Quantisation
- 4-bit granite-4.0-3b-vision-4bit
- Memory needed
- 6 GB Estimate, unified memory
- Download
- 3.0 GB Weights, estimated
- Downloads
- 120 This build, last 30 days
- Task
- Vision and text Text, Vision
- MLX builds
- 7 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 6 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-4.0-3b-vision-4bit --image photo.jpg --prompt "Describe this image"MLX builds
7 on mlx-community| granite-4.0-3b-vision-4bit | 4-bit | 3.0 GB | 6 GB | 120 |
|---|---|---|---|---|
| granite-4.0-3b-vision-mxfp4 | MXFP4 | 3.0 GB | 6 GB | 54 |
| granite-4.0-3b-vision-nvfp4 | 4-bit | 3.0 GB | 6 GB | 53 |
| granite-4.0-3b-vision-5bit | 5-bit | 3.5 GB | 6 GB | 62 |
| granite-4.0-3b-vision-6bit | 6-bit | 4.0 GB | 7 GB | 38 |
| granite-4.0-3b-vision-8bit | 8-bit | 4.9 GB | 8 GB | 58 |
| granite-4.0-3b-vision-mxfp8 | MXFP8 | 4.9 GB | 8 GB | 33 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.