Microsoft Research has introduced Fara1.5, a new family of browser computer-use agents available in 4B, 9B, and 27B parameter sizes. These agents, built on Qwen3.5 checkpoints and integrated with Microsoft's MagenticLite browser interface, demonstrated superior performance on the Online-Mind2Web benchmark compared to existing models like OpenAI's Operator and Google's Gemini 2.5 Computer Use. The training process involved a diverse mix of real web data and synthetically generated trajectories from a custom pipeline called FaraGen1.5, which includes cloned environments for gated tasks. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Sets a new benchmark for browser automation agents, potentially impacting how users interact with web services and how AI agents are developed.
RANK_REASON New product release from a major tech company (Microsoft) with performance claims against leading competitors. [lever_c_demoted from significant: ic=1 ai=1.0]
