TL;DR — Key Takeaways
- OpenAI introduced GPT-6.1 Terra with improved computer-use and coding capabilities, available through its API, ChatGPT Work and Codex.
- Standard API prices dropped to $1 per million input tokens and $5 per million output tokens, reductions of 50% and approximately 58%, respectively.
- OpenAI and independent benchmarking firm Artificial Analysis reported improved performance and lower benchmark costs, although actual enterprise savings will depend on individual workloads.
OpenAI has introduced GPT-6.1 Terra, an update to its mid-tier AI model designed to handle computer-based work at a lower cost. The company reported improvements in the model’s visual reasoning, application control and coding. Its standard API price for input tokens is half that of its predecessor, GPT-5.6 Terra, which could lower expenses for developers running large numbers of agentic tasks.
Terra sits between OpenAI’s higher-capability Sol and Astra models and its lowest-cost Luna tier. OpenAI says Terra can handle a range of workplace tasks, including drafting proposals, interpreting technical drawings, completing forms and editing videos. Developers can access GPT-6.1 Terra through the OpenAI API, which supports computer use, function calling, file search and connections to external tools through the Responses API. The model is also available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu subscribers.
Standard API pricing is $1 per million input tokens, five cents per million cached input tokens and $5 per million output tokens. In July, OpenAI reduced GPT-5.6 Terra’s prices to $2 per million input tokens and $12 per million output tokens. The new rates represent reductions of 50% and approximately 58%, respectively.
OpenAI tested Terra’s computer-use capabilities on Agents’ Last Exam, a benchmark involving complex professional tasks performed through software. Examples include converting architectural drawings into a 3D model and reviewing animated footage to produce a time-coded restoration log. At medium reasoning effort, GPT-6.1 Terra scored 15 percentage points higher than its predecessor at roughly one-third the cost per task. OpenAI also reported that Terra completed the benchmark’s tasks more than three times faster than its more powerful Astra and Sol models.
The model also performed better on Terminal-Bench 4.0, which evaluates how well coding agents solve problems using command-line tools. At maximum reasoning effort, GPT-6.1 Terra completed 50.3% of the tasks, more than double its predecessor’s previously published score of 21.5%, according to OpenAI, though the company said Sol and Astra remain better choices for the most complex coding tasks.
Independent AI benchmarking firm Artificial Analysis reported similar improvements in its own evaluation of GPT-6.1 Terra. The model scored six points higher than its predecessor on the firm’s Intelligence Index at maximum reasoning effort, while the average cost per task in that evaluation fell from $1.40 to $0.40.
However, lower costs in benchmark testing won’t necessarily translate into comparable savings for all enterprise workloads. In a September report on enterprise AI spending, Futurum analysts Brendan Burke and Daniel Newman recommended routing routine AI requests to less expensive models while measuring the cost of successfully completed tasks rather than token prices alone. But if Terra can reliably handle work that enterprises would otherwise assign to Sol or Astra, the savings could extend beyond cheaper API calls to the total cost of a workload. Read more about the updated Terra here.
