4-bit Open on Hugging Face
Parameters
2.5B
Weights (est.)
≈ 2.2 GB
Mac memory (est.)
8 GB+
Size is estimated from the parameter count and quantisation. Memory allows a fifth extra for context plus 2 GB for the system; long contexts need more.
Run it
With mlx-vlm on an Apple silicon Mac:
pip install mlx-vlmpython -m mlx_vlm.generate --model mlx-community/North-Micro-Vision-Instruct-4bit --image photo.jpg --prompt "Describe this image"
Details
- Base model
- CohereLabs/North-Micro-Vision-Instruct (quantized) · North Micro Vision Instruct on the radar →
- Vendor
- Cohere
- Task
- Vision + text
- Quantisation
- 4-bit
- Licence
- apache-2.0
- Downloads
- 76 in the last 30 days · 533 all time · 1 likes
- Published
- 2026-08-12 · updated 2026-08-12 (3mo ago)
- Languages
- en
- Tags
- cohere_compass, mlx-vlm, vision-language, multimodal, cohere, north, quantized
Other MLX builds of CohereLabs/North-Micro-Vision-Instruct
| Build | Quantisation | Weights | Mac memory | Downloads |
|---|---|---|---|---|
| North-Micro-Vision-Instruct-4bit (this one) | 4-bit | ≈ 2.2 GB | 8 GB+ | 76 |
| North-Micro-Vision-Instruct-mxfp4 | MXFP4 | ≈ 2.2 GB | 8 GB+ | 40 |
| North-Micro-Vision-Instruct-nvfp4 | 4-bit | ≈ 2.2 GB | 8 GB+ | 39 |
| North-Micro-Vision-Instruct-5bit | 5-bit | ≈ 2.4 GB | 8 GB+ | 36 |
| North-Micro-Vision-Instruct-6bit | 6-bit | ≈ 2.7 GB | 8 GB+ | 38 |
| North-Micro-Vision-Instruct-8bit | 8-bit | ≈ 3.1 GB | 8 GB+ | 49 |
| North-Micro-Vision-Instruct-mxfp8 | MXFP8 | ≈ 3.1 GB | 8 GB+ | 39 |
| North-Micro-Vision-Instruct-bf16 | bf16 | ≈ 5.0 GB | 8 GB+ | 43 |