North Micro Vision
#12 en Modèles Small & EdgeCohere · micro vision · depuis August 2026 · 4× · vu le 13 août 2026
40
Momentum
North Micro Vision is an open-weight vision-language model developed by Cohere (Cohere Labs) with 2.4 billion parameters, supporting native-resolution image input and designed specifically for document understanding, OCR, charts, and multilingual visual tasks. It is released under the Apache 2.0 license as open weights on Hugging Face and is intended as a compact foundation for fine-tuning and edge/local deployments. The language backbone supports a 128K-token context window, though the validated multimodal operating range is up to 8K tokens.
Historique du momentum
15.05.13.08.
Fonctionnalités
| Context Window (Tokens) | 128,000 tokens (language backbone); validated multimodal range up to 8,000 tokens |
| License | Apache 2.0 (open model weights) |
| Multimodality | Vision-language model with native-resolution support; capabilities in VQA, captioning, grounding, OCR, charts, and documents, multilingual and multi-image capable |
| Platform | Open-weight model, available on Hugging Face (CohereLabs/North-Micro-Vision-Instruct), with NVIDIA AutoModel and Axolotl fine-tuning support |
| Release Date | August 2026 |