MLX Community · First MLX build 22 Dec 2024
deepseek-vl2-small
MLX
TextVision deepseek
A 16B parameter vision and text model, with 5 MLX builds, from 3-bit to bf16.
- Parameters
- 16B From Hugging Face
- Quantisation
- 4-bit deepseek-vl2-small-4bit
- Memory needed
- 13 GB Estimate, unified memory
- Download
- 9.3 GB Weights, estimated
- Downloads
- 138 This build, last 30 days
- Task
- Vision and text Text, Vision
- MLX builds
- 5 On mlx-community
- Devices
- Mac Apple silicon
Memory
Will it fit?Needs about 13 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/deepseek-vl2-small-4bit --image photo.jpg --prompt "Describe this image"MLX builds
5 on mlx-community| deepseek-vl2-small-3bit | 3-bit | 7.3 GB | 11 GB | 74 |
|---|---|---|---|---|
| deepseek-vl2-small-4bit | 4-bit | 9.3 GB | 13 GB | 138 |
| deepseek-vl2-small-6bit | 6-bit | 13.3 GB | 17 GB | 81 |
| deepseek-vl2-small-8bit | 8-bit | 17.3 GB | 22 GB | 68 |
| deepseek-vl2-small-bf16 | bf16 | 32.3 GB | 39 GB | 84 |
Sizes are estimated from Hugging Face’s parameter counts and the quantisation. Downloads are for the last 30 days.