OpenAI Launches GPT-6.1 Terra for Computer Use

OpenAI has introduced GPT-6.1 Terra, a model that the company says combines visual reasoning, computer use, and natural writing. The launch targets visual and computer-based work such as drafting proposals, interpreting technical drawings, filling out forms, and editing videos.

Standard API pricing for the model is $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens. These rates are 50% lower than standard API prices for GPT-5.6 Terra and GPT-6.1 Sol. The model is available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users, as well as through the OpenAI API under the name gpt-6.1-terra.

According to OpenAI's evaluations on Agents' Last Exam, a benchmark for professional tasks, GPT-6.1 Terra scored 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort while running at about one-third the cost. OpenAI says it also matched the performance of GPT-6 Astra at one-tenth of its price per token, while completing tasks over three times faster than both GPT-6 Astra and GPT-6.1 Sol. At max effort, the company reports that Terra operated at over three times the token efficiency of Claude Opus 5.5 at one-fourth of its price per token, yielding a ~9x lower cost per task.

On MMMU-Pro visual reasoning tests, OpenAI says GPT-6.1 Terra scored higher than GPT-5.6 Terra across every reasoning setting. On OSWorld 2.1, which tests desktop application usage, the company reports that the model surpassed the top score of GPT-5.6 Terra at medium effort.

OpenAI also reported reductions in factual errors on internal benchmarks. According to the company, at low reasoning effort, error rates fell from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra, a drop of nearly 37 percentage points. At maximum reasoning effort, OpenAI states the error rate fell by 18 percentage points to 4.3%.

For software engineering tasks evaluated on Terminal-Bench 4.0, OpenAI says GPT-6.1 Terra solved 50.3% of command-line coding problems at maximum reasoning effort, compared to 21.5% for GPT-5.6 Terra. On Terminal-Bench-Science 0.1, the company reports that the model finished within five points of Opus 5.5 at max reasoning.

Safety testing under OpenAI's Preparedness Framework placed GPT-6.1 Terra at the Critical threshold in cybersecurity.