OpenAI Unveils GPT-6.1 Terra with Lower API Prices and Computer Use

OpenAI introduced GPT-6.1 Terra on October 26, marking a strategic shift toward reducing the cost of completing automated tasks rather than simply pushing the absolute ceiling of model capabilities. The newly unveiled model is designed to handle high-frequency, repetitive daily office work, such as reading technical drawings, filling out forms, editing videos, and operating desktop software.

Standard application programming interface (API) prices for GPT-6.1 Terra are 50% lower than those of the previous-generation GPT-5.6 Terra. The pricing is set at $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens. Compared directly to OpenAI's flagship model, GPT-6 Astra, both input and output unit prices for Terra are just one-tenth of the flagship rates.

OpenAI positioned the model as an efficient tool for routine tasks that require computer use, which involves an artificial intelligence system perceiving a screen and interacting with applications through clicking and typing similarly to a human operator. The model is currently rolling out to Plus, Pro, Enterprise, and Edu users within ChatGPT Work and Codex, while developers can access it via the API under the identifier gpt-6.1-terra.

Performance benchmarks published by OpenAI indicate that GPT-6.1 Terra closes the performance gap with flagship products while maintaining a significantly lower price point. On the Agents' Last Exam test, which evaluates software utilization for professional tasks, the model scored 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort, bringing its performance on par with GPT-6 Astra while reducing per-task costs to roughly one-third of its predecessor.

In addition to task efficiency, OpenAI highlighted improvements in factual accuracy and hallucination control. Internal high-difficulty tests demonstrated that the error rate for GPT-6.1 Terra dropped to 16.7% at low reasoning effort—down from 53.6% for GPT-5.6 Terra—and fell to 4.3% at maximum reasoning effort. Third-party evaluator Artificial Analysis noted that the model's tendency to guess when facing uncertain questions decreased markedly under private benchmarking.

Despite these enhancements, OpenAI drew clear operational boundaries for its product lineup. The company stated explicitly that GPT-6.1 Sol and GPT-6 Astra remain the appropriate choices for the most complex software engineering projects and deep research tasks. Industry analysts observe that as screen-understanding and software-operation costs decline, the economic threshold for deploying automated agents across large-scale daily office environments continues to fall.