OpenAI has released GPT-6.1 Terra, a model the company says is optimized for visual reasoning and “computer use” — the ability to read screens and interact with software to perform routine office tasks. OpenAI positioned Terra between its higher-end GPT-6 models and smaller offerings, saying it targets workflows such as drafting proposals, filling forms, interpreting technical drawings and processing invoices.
What OpenAI announced
OpenAI reported that GPT-6.1 Terra improves accuracy over the model it replaces and offers large cost and speed advantages in internal benchmarks. On a workplace-oriented test called Agents’ Last Exam, OpenAI said Terra at medium reasoning effort scored 15 percentage points higher than GPT-5.6 Terra while operating at roughly one-third the cost. The company also said Terra matches GPT-6 Astra on that test at about one-tenth the price per token.
“At max effort, Terra is over three times as token-efficient as Claude Opus 5.5, and one-fourth its price per token, resulting in a ~9x lower cost per task.” — OpenAI tweet announcing GPT-6.1 Terra
OpenAI said Terra’s error rate fell sharply on an internal benchmark designed to provoke mistakes: at low effort the error rate dropped to 16.7% from 53.6% for GPT-5.6 Terra, and at maximum effort it reached 4.3%, a reduction OpenAI said nearly matches results from GPT-6 Astra and GPT-6.1 Sol. In another internal test, Terminal-Bench 4.0, Terra solved 50.3% of tasks at maximum effort versus 21.5% for GPT-5.6 Terra.
Performance, cost and recommended use
OpenAI highlighted three commercial advantages for Terra:
- Speed: the company said Terra completes tasks more than three times faster on average than GPT-6 Astra and GPT-6.1 Sol.
- Token efficiency: OpenAI reported significant gains versus prior Terra and compared efficiency favorably against Anthropic’s Claude Opus 5.5 in its own tests.
- Lower cost per task: OpenAI said the combination of efficiency and pricing yields substantially lower cost per task for many routine workflows.
OpenAI noted that for the most complex coding work it still recommends higher-capability models such as GPT-6.1 Sol and GPT-6 Astra.
Availability
According to the announcement, GPT-6.1 Terra is available to ChatGPT users on paid tiers and through OpenAI’s API offerings. OpenAI released benchmark comparisons and cost figures drawn from its internal testing to illustrate the model’s target use cases and relative economics.
The company’s published numbers and tweets frame Terra as a mid-tier option intended to automate repetitive, high-volume tasks where speed and lower operating cost matter.
