Vision-language models
3 stories. An explainer is being written.
Vision-language models in the news
·NVIDIA
Nvidia expands Nemotron family with Super VL and two Bolt models
Open weights now cover servers down to 4GB edge devices
·Hugging Face Hub
OpenBMB ships quantized MiniCPM-V 4.7, its first MoE vision model
A 35-billion-parameter model that runs at 13+ tokens a second on an iPhone
·arXiv cs.LG
Study: harmful-meme failures in VLMs are a routing problem
Researchers find Gemma-3 and Qwen3.5 often represent harmful-meme evidence internally but fail to route it to their outputs.