Microsoft Research AI Frontiers has released Fara1.5-27B, a multimodal computer use agent designed for web browsers. The model observes browser interfaces through screenshots and executes tasks by emitting structured tool calls such as clicking, typing, scrolling, and searching.
- Supervised fine-tuned from Qwen3.5-27B using data generated by the FaraGen1.5 multi-agent pipeline.
- Vision-only perception that tracks internal reasoning and trajectory history as text to predict grounded actions.
- Co-designed with MagenticLite for recommended deployment in research and production environments.
- Available in 4B, 9B, and 27B parameter sizes on Hugging Face.
The agent automates repetitive web tasks like form filling and booking while serving as a grounding model for other agents requiring pixel-accurate action prediction.