Microsoft ships Fara1.5-27B, a screenshot-based browser agent for web tasks
pmttyji · reddit · 2026-07-23
Microsoft releases Fara1.5-27B, a browser computer-use agent
Microsoft Research AI Frontiers introduced Fara1.5-27B, a multimodal computer-use agent for web browsers.
- It works from screenshots only at perception time, not the DOM or accessibility tree.
- The model emits structured tool calls such as click, type, scroll, visit URL, and web search to complete tasks end to end.
- It is supervised fine-tuned from Qwen3.5-27B using data generated by FaraGen1.5, a multi-agent pipeline that synthesizes, executes, and verifies web trajectories.
- Microsoft says it is co-designed with MagenticLite, which is the recommended deployment setup.
- The model is intended for repetitive browser workflows such as form filling, booking, shopping, and information seeking, but Microsoft warns against high-stakes or sensitive uses.
The release also lists key limitations: vision-only perception can be fooled by deceptive rendering or prompt injection, long trajectories accumulate errors, and run-to-run variance remains significant.
More from coding & agent
- 'My intelligence stack': developer splits AI coding work across Opus 5.5, Fable and Astra with human review gates — sh_reya · 2026-10-27
- Baseten webinar: evaluate open models for coding agents by benchmarking on your own merged PRs — baseten · 2026-10-27
- Thorsten Ball on working with agents: run more evals, automate everything, use the best models — omarsar0 · 2026-10-27
- DeepMind paper: LLMs can't sample randomly on their own — pass random numbers from code instead — rohanpaul_ai · 2026-10-27
- BELL: Open-Source Stack Combines Browserbase, Eve, iMessage and Stripe Link for Personal Agents — jeff_weinstein · 2026-10-27
- Legora runs 2,000 parallel agent evals on Modal Sandboxes for legal AI — blelbach · 2026-10-27