Microsoft releases Fara-7B, a 7-billion parameter small language model designed for autonomous computer use through visual perception and direct interaction with web interfaces. The model achieves state-of-the-art performance in its size class across multiple web agent benchmarks, completing tasks in ~16 steps versus ~41 for comparable models. Trained on 145K synthetic trajectories using the Magentic-One framework, Fara-7B can automate web tasks like booking travel, shopping, and form filling by directly predicting mouse and keyboard actions. The release includes WebTailBench, a new benchmark with 609 real-world tasks, and supports both Azure Foundry hosting and self-hosted VLLM deployment.