Microsoft releases Mage-VL codec-native multimodal foundation model
July 28, 2026
Mage-VL provides proactive-streaming capabilities for image and video understanding using a codec-native architecture. The model is now available as an open-weights release on Hugging Face.
HOW THIS AFFECTS YOU
●
builderYou can integrate proactive-streaming multimodal capabilities into video-based applications via Hugging Face.
●
researcherThe codec-native approach offers a new architecture for studying streaming multimodal foundation models.