OpenAI launches GPT-6.1 Terra with API prices 50% lower than its two predecessors
Key Facts
OpenAI announced the launch of GPT-6.1 Terra, a model it said combines detailed visual reasoning and precise computer use with clearer, more natural writing. It has been available since October 26 to subscribers on the Plus, Pro, Enterprise and Edu plans in ChatGPT Work and Codex, and to developers through the company’s API under the identifier gpt-6.1-terra. The company has put lower usage costs at the center of the release, saying standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol. The launch comes with OpenAI presenting performance comparisons in specific tests, as well as classifying the model’s capabilities as critical in cybersecurity. The company says that classification is accompanied by the safeguards it developed for GPT-6 Astra.
In Agents’ Last Exam, at medium reasoning effort, Terra scored 15 percentage points higher than GPT-5.6 Terra, according to OpenAI, at a cost of about one-third that of the previous model. The company added that its performance in the test was comparable to GPT-6 Astra, at one-tenth the token price of that model. This comparison links performance and cost in the specified test and does not offer a general judgment on every type of use. OpenAI also says the same plans allow users to complete more tasks in ChatGPT Work and Codex with Terra than with GPT-5.6 Terra, and far more than with GPT-6.1 Sol and GPT-6 Astra. According to the company, that means the same usage limits cover a larger number of tasks with the new model.
API documentation sets the standard price at $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. In a cost comparison at maximum reasoning effort, OpenAI said Terra completed tasks with more than three times the token efficiency of Claude Opus 5.5, while its token price was one-quarter that of the model. It said those differences resulted in a task cost about 9 times lower in the same comparison. These ratios remain tied to the effort level and comparison specified by the company, while published prices reflect standard API rates. With the launch, OpenAI therefore presented separate figures for token prices and task costs, alongside the announced reduction in API rates compared with its two predecessors.
In Terminal-Bench 4.0, Terra solved 50.3% of tasks at maximum reasoning effort, compared with a published score of 21.5% for GPT-5.6 Terra, according to OpenAI. The company described the result as more than double the previous model’s published score in the test. The percentage does not apply to all coding or computer-use tasks, but to the two models’ results in this test at the stated effort level. The Agents’ Last Exam comparison is based on medium reasoning effort, while Terminal-Bench reports the result at maximum effort, giving each figure a different context. With these results, the company extended its announcement beyond the model’s features to specific performance and cost measures in separate tests.
On safety, OpenAI said it treats Terra’s capabilities as critical in cybersecurity and high in the biological and chemical domains under its preparedness framework. It added that the model retains the safeguards it developed for GPT-6 Astra, while using the same set of safeguards as GPT-6.1 Sol and GPT-6 Astra. An appendix to Terra’s safety card was published on October 26, 2026, the day the model was announced. This classification places the model’s capabilities and safeguards among the launch’s main details, alongside prices and performance tests. The launch announcement did not specify a later date for additional access; the company said access for subscribers and developers began on the day of the announcement.