Fara-7B is an agentic model designed for computer use tasks, enabling AI systems to interact with user interfaces, browsers, and applications. The repository includes an agent harness for running the model in autonomous workflows and evaluation frameworks including WebTailBench (609-task benchmark suite) and CUAVerifierBench (human-annotated dataset for evaluating CUA verifiers). The model can be accessed via Azure Foundry or Hugging Face.