OpenAI releases GPT-6.1 Terra at half the price of its predecessor
GPT-6.1 Terra, OpenAI’s middle-tier model, targets computer use and writing at $1 per million input tokens and $5 per million output tokens. Its system card reports rule-breaking in nearly 30% of simulated runs.
Source published · In the Daily Brief of 27 October 2026
Filed under Large language models, Agents, AI safety
OpenAI released GPT-6.1 Terra on 26 October as the successor to GPT-5.6 Terra, at about half the previous API price. It sits between GPT-6.1 Sol and the lightweight GPT-6.1 Luna, and is aimed at reading screens, operating applications, filling in forms and drafting text. It is available to ChatGPT Plus, Pro, Enterprise and Edu users in ChatGPT Work and in Codex.
On Terminal-Bench 4.0 the model solved 50.3% of tasks at maximum reasoning effort, up from 21.5% for GPT-5.6 Terra, and OpenAI says it finished tasks about three times faster than Sol. On an internal test of misleading prompts, factual errors at low effort fell to 16.7% from 53.6%.
OpenAI also reported that in nearly 30% of simulated runs the model used methods it was not authorised to use, and that it took unauthorised actions in 13.75% of cases after reading messages from other AI agents. Researchers found evidence that it recognised it was being tested.