OpenAI launches GPT-6.1 Terra for computer use

OpenAI has launched GPT-6.1 Terra, a model the company describes as bringing fast and capable computer use to a broad audience. The announcement focuses on AI agents that can interact with software interfaces directly, clicking through menus and carrying out tasks instead of only answering questions in text. If the claim holds up, the model could expand automation into everyday work and consumer activities that depend on existing applications rather than custom integrations. Computer use is an important frontier because much of the digital economy still runs through interfaces designed for people. Agents that can operate those interfaces could help with data entry, research, scheduling, customer support and back-office workflows. They could also make automation available to smaller organizations that cannot build bespoke software connections. OpenAI's pitch suggests Terra is meant to make this kind of agentic behavior faster and more accessible, rather than limited to specialized pilots. However, the launch came with few technical details. OpenAI did not provide enough public information in the brief announcement to verify Terra's speed, reliability, error rates or safety properties. Independent evaluations will be needed, especially for computer-use agents, where mistakes can have real consequences. An agent that misclicks, misreads a screen or takes an unintended action can create problems that a chatbot does not. Customers will want to know how the model handles permissions, sensitive data and recovery from failure. The release also raises the competitive bar. Agentic AI has become a major focus for model providers, and each new system invites comparison on task completion, cost and safety. Terra's success will depend less on a flashy demo than on consistent performance across messy real-world software. If OpenAI can deliver that, GPT-6.1 Terra could become an important step toward practical computer-use agents. If not, it will add to the debate over how quickly autonomous software operation can be trusted in ordinary business settings.