Microsoft Releases Open-Weight AI Model Fara1.5, Scores 72% on Web Browsing Benchmark, Beating OpenAI and Google

Microsoft Research released open-weight AI model Fara1.5 this week, with the 27-billion-parameter version scoring 72% on the Online-Mind2Web benchmark, surpassing OpenAI Operator (58.3%) and Google Gemini 2.5 Computer Use (57.3%). The model family comes in three sizes—4 billion, 9 billion, and 27 billion parameters—built on Alibaba's Qwen3.5 base model and fine-tuned by Microsoft for web browsing tasks. All weights are publicly released.

Fara1.5-9B is now live on Azure AI Foundry, with the 4-billion and 27-billion variants arriving shortly. Unlike OpenAI's and Google's proprietary cloud-based solutions, Microsoft's open approach allows users to run the model on their own hardware without subscription fees.