GPT-6.1 Terra: API Pricing, Tools and Availability

OpenAI introduced GPT-6.1 Terra on October 26, 2026, as a lower-priced model built for visual reasoning, computer use and natural writing. For developers and teams, the useful questions are what the API actually costs, which inputs and tools work, and whether the model is accessible in their product. Those practical limits matter more than assuming vendor benchmark results will transfer unchanged to a production workflow.

API price: standard rates and exceptions

The API model ID is gpt-6.1-terra. OpenAI lists a 1,050,000-token context window and up to 128,000 output tokens. Do not treat the headline price as a universal flat rate: prompts with more than 272,000 input tokens are billed at 2x the input and cache rates and 1.5x the output rate for the full request. Extended cache retention, tool usage, regional processing and faster modes can add charges. Check the official model reference before pricing long-document or agent workflows.

Inputs, tools and computer use

The official model reference lists text input and output and image input, but no native audio or video modalities. It supports structured outputs, function calling, computer use, web search, file search and code interpreter. OpenAI specifies that tool calling requires the Responses API. A model using an application interface to edit media is different from directly accepting a video stream as a native input. Developers should validate the exact input and tool path required by their product.

Availability in ChatGPT Work, Codex and Copilot

OpenAI says Terra is available through its API and in ChatGPT Work and Codex for eligible Plus, Pro, Enterprise and Edu users. That statement does not establish universal availability in ordinary ChatGPT conversations or every region. GitHub separately says Terra is generally available to Copilot Pro, Pro+, Max, Business and Enterprise subscribers across supported editors, the CLI, cloud agent, app, web and mobile interfaces. GitHub also warns rollout is gradual, and business administrators can control model access through Copilot settings.

How to test it in a real workflow

1. Pick one repeatable, low-risk task with a measurable success criterion, such as extracting invoice fields or matching catalog images.
2. Confirm the API model and tool route you need, including Responses API requirements for tool calling.
3. Record ordinary, cached and output tokens, tool calls, retries and any long-context price multipliers.
4. Review errors and permission boundaries; require approval before actions that alter external records.
5. Compare accuracy, elapsed time, review effort and total cost against your current model on the same examples.

As a simple token-only illustration, 100,000 ordinary input tokens and 10,000 output tokens cost about $0.15 at the listed standard rates before additional charges. This arithmetic is not a prediction of the total cost of a real agent run. Tools, retries, caching behavior and long contexts can change the result. A pilot should measure end-to-end cost per successfully completed task, not only the advertised price per token.