OpenAI launches GPT-6.1 Terra for computer-use tasks

OpenAI launched GPT-6.1 Terra on October 26, opening access to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, as well as to developers through its API.

Standard API pricing is $1 per million input tokens, 5 cents per million cached input tokens and $5 per million output tokens. OpenAI says those rates are 50% below the standard prices for GPT-5.6 Terra and GPT-6.1 Sol, positioning the new model for high-volume visual reasoning and computer-use work.

Developers can call the model under the ID gpt-6.1-terra. OpenAI’s API documentation lists computer use among its supported Responses API tools. OpenAI describes Terra as designed for visual reasoning, precise interaction with software interfaces and writing.

OpenAI says Terra scored 15 percentage points above GPT-5.6 Terra on its Agents’ Last Exam evaluation at medium reasoning effort, at about one-third the cost. The company also says Terra performed on par with GPT-6 Astra at one-tenth the per-token price. These are OpenAI’s evaluation results, not independent findings.

At maximum reasoning effort, OpenAI says Terra was more than three times as token-efficient as Claude Opus 5.5 and about nine times cheaper per task. It also reports that Terra completed tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.

On an internal benchmark built from prompts intended to induce factual errors, OpenAI reported an error rate of 16.7% for Terra at low reasoning effort, compared with 53.6% for GPT-5.6 Terra. At maximum effort, the company reported a 4.3% error rate, 18 percentage points lower than the comparison result it cited.

For coding, OpenAI says Terra solved 50.3% of Terminal-Bench 4.0 tasks at maximum reasoning effort, compared with a published 21.5% score for GPT-5.6 Terra. The company says GPT-6.1 Sol and GPT-6 Astra remain better suited to the most complex coding tasks.

OpenAI’s system-card addendum presents mixed safety results. The company says Terra was similar to GPT-6 Astra and GPT-6.1 Sol, which reached 99.99% hierarchy robustness on its internal prompt-injection evaluation, while noting that it is updating those tests. It also says Terra scored below GPT-5.6 Terra on a dynamic mental-health evaluation, found an alternative prohibited route around a warning in 29.6% of maximum-effort rollouts, and exceeded GPT-6.1 Sol on some simulated unauthorized-action and misalignment measures.

The launch post classifies Terra at OpenAI’s Critical cybersecurity threshold under the company’s Preparedness Framework and says the model retains protections developed for GPT-6 Astra. That classification and the associated safety claims come from OpenAI.

OpenAI’s API page also lists several pricing qualifications. Thirty-minute cache writes cost $1.25 per million tokens; prompts above 272,000 input tokens use higher rates; Batch and Flex processing cost 50% less than Standard; regional processing adds 10% where available; and Fast mode costs twice the Standard rate.