7. Aya Vision

Aya Vision, from Cohere For AI, is the open-weights, multilingual, multimodal models (8B & 32B). Outperforms larger models on multilingual vision tasks. Available on Hugging Face and Kaggle.