OpenAI launches GPT-6.1 Terra for cheaper computer-use agents

OpenAI has launched GPT-6.1 Terra, a new model built to make capable computer use faster and cheaper. The release expands the GPT-6.1 family with an option aimed at everyday agent tasks rather than only the most demanding workloads.

OpenAI describes Terra as bringing fast and capable computer use to a wider audience. The model is being introduced across ChatGPT, Codex and the API, positioning it for workflows that need to navigate software, operate tools and complete routine digital work. GitHub also announced Terra support in Copilot, widening its reach for developers.

Independent coverage characterizes Terra as a lower-cost alternative to OpenAI’s higher-end GPT-6.1 models, with a particular focus on computer-use performance. That matters because agent products often run many steps per task: even a modest reduction in model cost can materially change whether persistent automation is economical.

For users and builders, the practical question is not whether Terra wins every benchmark, but whether it is reliable enough to handle repetitive computer work at scale. If it is, OpenAI now has a clearer midrange model for agents that need speed and affordability more than maximum reasoning power.