A GPT that sees the screen and clicks for you
OpenAI released GPT-6.1 Terra, a fast model that is strong at understanding images and screens and operating computers. It is available in ChatGPT Work and Codex on Plus and higher plans, and in the API (gpt-6.1-terra) at $1 input and $5 output per 1M tokens.
Key facts
| Plans | Plus, Pro, Enterprise, Edu |
|---|---|
| Where to use | ChatGPT Work, Codex, API |
| API name | gpt-6.1-terra |
| API input | $1 per 1M tokens ($0.05 cached) |
| API output | $5 per 1M tokens |
| Previous model | GPT-5.6 Terra |
What happened
OpenAI introduced GPT-6.1 Terra in its official announcement. It is a fast model built to understand screens and images, operate computers and write well. OpenAI gives examples such as drafting proposals, interpreting technical drawings, filling in forms and editing video.
Why it matters
Screen-based work comes in large volumes, so cost is the key issue. Terra's standard API price is 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. OpenAI gives the example of high-volume work such as reading received invoices and entering them into accounting software.
What changed
All figures below were measured by OpenAI itself and have not been reproduced by this site.
- Agents' Last Exam (medium reasoning): about one-third the cost of GPT-5.6 Terra, with a score 15 percentage points higher.
- On the same exam it is on par with GPT-6 Astra at one-tenth the price per token. OpenAI says it finishes tasks more than three times faster than Astra and Sol.
- Error-inducing benchmark (maximum reasoning): the error rate fell 18 percentage points to 4.3%.
Who uses it, and where
- Plus, Pro, Enterprise and Edu users can use it in ChatGPT Work and Codex.
- Developers can use it in the API as
gpt-6.1-terra. - OpenAI says it can handle more work than GPT-5.6 Terra within the same usage limits.
- For the most complex coding, OpenAI itself recommends GPT-6.1 Sol and GPT-6 Astra.
Cost and safety
- The API costs $1 input, $0.05 cached input and $5 output per 1M tokens.
- OpenAI stated that the model reached the Critical threshold for cybersecurity in its preparedness evaluation, and that it is keeping the safeguards used for GPT-6 Astra.
- When letting it operate a computer directly, it is safer to have a person confirm actions such as payments, deletions and sending.
Try it
One approach is to pick a task you currently repeat by hand. Give the same input to your current model and to Terra and compare the results, time taken and usage; that tells you more about your own work than benchmarks do.
Sources
Questions
Where can I use GPT-6.1 Terra?
In ChatGPT Work and Codex on Plus, Pro, Enterprise and Edu, and in the API.
What is the API name and price?
It is gpt-6.1-terra, priced per 1M tokens at $1 input, $0.05 cached input and $5 output.
Is it also the best for coding?
OpenAI says GPT-6.1 Sol and GPT-6 Astra are better for the most complex coding tasks.
