North Micro Vision Instruct
North Micro Vision Instruct is an open-weights language model from Cohere, released August 10, 2026. It has 2.5B parameters (dense). The context window is 488K tokens.
Check it against your own PC →
Can you run North Micro Vision Instruct locally?
The smallest common GPU it runs on is 6 GB of VRAM (with 32 GB of RAM): Q6_K, fully on GPU, roughly 54 tokens per second. Estimates assume 32 GB of DDR5 RAM, a 16K context and llama.cpp-style GGUF quantization. Enter your exact hardware in the live app.
| GPU memory | Verdict | Best quant | VRAM | RAM | tok/s |
|---|---|---|---|---|---|
| 6 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 3.5 GB | – | ~54 |
| 8 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 3.5 GB | – | ~74 |
| 12 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 3.5 GB | – | ~96 |
| 16 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 3.5 GB | – | ~130 |
| 24 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 3.5 GB | – | ~220 |
| 32 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 3.5 GB | – | ~350 |
Ready-made quantizations on Hugging Face: mlx-community/North-Micro-Vision-Instruct-4bit · OpenMed/North-Micro-Vision-Instruct-4bit-mlx · mlx-community/North-Micro-Vision-Instruct-8bit · OpenMed/North-Micro-Vision-Instruct-8bit-mlx · OpenMed/North-Micro-Vision-Instruct-6bit-mlx · mlx-community/North-Micro-Vision-Instruct-mxfp4
Benchmark results
Epoch AI has not published results for North Micro Vision Instruct yet. New models are usually evaluated within days to weeks.
Links
Hugging Face: CohereLabs/North-Micro-Vision-Instruct
More from Cohere
tiny aya en thinker · tiny aya l2 thinker · North Mini Code · Command A Plus · Tiny Aya Global · Tiny Aya Fire · Tiny Aya Water · Tiny Aya Earth