# Cohere Labs Ships a 2.4-Billion-Parameter Vision Model That Reads Images at Native Resolution

North Micro Vision pairs a 400-million-parameter SigLIP2-derived encoder with a 2-billion-parameter language model and is released under the Apache 2.0 licence.

- Published: 2026-08-18T06:24:23.025Z
- Canonical: https://polylog.news/ai/2026-08-18/cohere-labs-ships-a-2-4-billion-parameter-vision-model-that
- Publisher: Polylog (AI desk)
- Section: tech
- Sources: [Polylog editors](https://polylog.news), [Cohere Labs (Hugging Face)](https://huggingface.co/blog/CohereLabs/meet-north-micro-vision-instruct), [NVIDIA NeMo AutoModel](https://docs.nvidia.com/nemo/automodel/model-coverage/vision-language-models/coherelabs/north-micro-vision)

Cohere Labs released North Micro Vision Instruct, the smallest vision-language model in its lineup at 2.4 billion parameters, under the Apache 2.0 licence. The company's release writeup describes a 400-million-parameter vision encoder built…

This story is for subscribers. Read it in full at https://polylog.news/ai/2026-08-18/cohere-labs-ships-a-2-4-billion-parameter-vision-model-that (subscription information: https://polylog.news/pricing).