microsoft/Fara1.5-9B
Fara1.5-9B is a 9-billion parameter multimodal computer use agent (CUA) developed by Microsoft Research AI Frontiers, designed for web browsers. It operates vision-only, perceiving the browser through screenshots to complete end-to-end web tasks by emitting structured tool calls. With a 262,144-token context length, it excels at automating repetitive web actions like form filling, bookings, and shopping, while incorporating critical-points safety for sensitive operations.
Loading preview...
Fara1.5-9B: Web Computer Use Agent
Fara1.5-9B, developed by Microsoft Research AI Frontiers, is a 9-billion parameter multimodal computer use agent (CUA) specifically designed for web browsers. It functions by observing the browser through screenshots, rather than the DOM, and generates structured tool calls (e.g., click, type, scroll, visit URL) to complete complex web tasks end-to-end. The model is supervised fine-tuned from Qwen3.5-9B using data synthesized by FaraGen1.5, a multi-agent pipeline.
Key Capabilities
- End-to-end web task completion: Automates tasks like form filling, reservations, job applications, and shopping.
- Vision-only perception: Operates solely on screenshots, mimicking human interaction without DOM access.
- Coordinate-grounded actions: Predicts pixel-level targets for clicks and drags directly.
- Critical-points safety: Trained to pause and seek user confirmation before irreversible actions (e.g., payments, submissions, personal info entry).
- Extended context: Features a 262,144-token context length, suitable for multi-screenshot trajectories and full action history.
Good For
- Automating repetitive web tasks in a sandboxed environment.
- Use cases requiring pixel-accurate action prediction on web interfaces.
- Research and production deployments with human-in-the-loop monitoring and strict access controls, ideally within MagenticLite.