
Salesforce AI, in collaboration with the University of Washington, has introduced xGen-MM, also known as BLIP-3, a new family of open-source large multimodal models (LMMs) with a 4b parameter size. This innovative framework aims to advance visual language understanding by integrating text and image comprehension. xGen-MM is designed with competitive performance among open-source models and includes curated large-scale datasets, training recipes, model architectures, and safety tuning. The release of xGen-MM marks a significant improvement over its predecessor, BLIP-2, offering enhanced training and performance capabilities. The framework has gained notable attention, ranking #1 on Hugging Face, a popular platform for AI models.







Salesforce AI Research Introduce xGen-MM (BLIP-3): A Scalable AI Framework for Advancing Large Multimodal Models with Enhanced Training and Performance Capabilities https://t.co/naBEoKKPSf #AI #Salesforce #MultimodalModels #ArtificialIntelligence #TechInnovation #ai #news #ll… https://t.co/q08PASJjpg
Salesforce AI Research Introduce xGen-MM (BLIP-3): A Scalable AI Framework for Advancing Large Multimodal Models with Enhanced Training and Performance Capabilities Researchers from Salesforce AI Research and the University of Washington have introduced the xGen-MM (BLIP-3)… https://t.co/R2lD412mVz
[CV] xGen-MM (BLIP-3): A Family of Open Large Multimodal Models https://t.co/C1E8sIAfai - xGen-MM (BLIP-3) is a family of Large Multimodal Models (LMMs), with curated datasets, training recipes, architectures, and resulting models. - xGen-MM improves upon BLIP-2 by… https://t.co/dhuuq0ym1U