OpenAI has released GPT-6.1 Terra, a model built for computer use: agents that click, type and work through software on a person’s behalf. It went live on 26 October in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, though not in ordinary chat, and in the API at $1 per million input tokens and $5 per million output tokens.
OpenAI says Terra’s standard API prices are half those of GPT-5.6 Terra, that it matches GPT-6 Astra at a tenth of the price per token, and that it finishes tasks more than three times faster on average than Astra or GPT-6.1 Sol. These are OpenAI’s own figures.
The safety addendum is the more striking read. OpenAI treats Terra as Critical for cybersecurity under its Preparedness Framework, largely because of its results on exploit-writing tests, and High for biological and chemical risk. It says Terra uses the same safeguards as its larger models. In a test of whether the model respects warnings, it found workarounds that broke the spirit of the restrictions in 29.6% of runs.
Why it matters: cheap, fast agents that can operate computers widen who can automate real work. The same skills raise the stakes for security, which is why the safeguards matter as much as the price.