Ai2 @ai2.bsky.social · 18/03/2026We're releasing three models, all open source: 🖼️MolmoPoint-8B—general-purpose pointing across images & video 🖥️MolmoPoint-GUI-8B—specialized for apps, websites, & software interfaces 🎥MolmoPoint-Vid-4B—optimized for counting & tracking in video 141
Ai2 @ai2.bsky.social · 18/03/2026We're also releasing 2 new open datasets: 📦MolmoPoint-GUISyn (used to train our GUI model)—36K high-res screenshots spanning desktop, web, and mobile, with 2M+ annotated points 📦MolmoPoint-TrackData (used to train our video model)—human-annotated & synthetic tracks with complex occlusion + motion 140
Ai2 @ai2.bsky.social · 18/03/2026VLMs already have visual tokens. Letting them point by selecting those tokens turns out to be simpler, faster, & better. 🤖 Models: huggingface.co/collections/... 📦 Data: huggingface.co/collections/... 💻 Code: github.com/allenai/molmo2 📖 Blog: allenai.org/blog/molmopo...huggingface.coMolmoPoint - a allenai CollectionMolmoPoint models 052
