Summary
- OpenAI released GPT-6.1 Terra on October 26, positioning it for visual reasoning, computer use and writing.
- Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens.
- OpenAI reports improvements on its evaluations, but benchmark results do not guarantee performance on an individual workflow.
What is GPT-6.1 Terra?
OpenAI released GPT-6.1 Terra on October 26, adding a model aimed at visual reasoning and computer-based work to its GPT-6 family. The company says it can interpret technical drawings, draft proposals, fill out forms and work across software interfaces. The API identifier is gpt-6.1-terra. [1] [2]
The distinction is not simply that the model can describe an image. OpenAI says Terra is designed to use information on a screen as part of a workflow, including acting in applications. Its documentation lists image and text input, text output and computer use through the Responses API; it does not list native video or audio input as supported modalities. [1] [3]
OpenAI says the model is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and to developers through its API. Its learning documentation also lists Business plans, while noting that workspace controls and setup requirements can affect access. The same documentation says Terra is not available in Chat. [1] [4]
How much does GPT-6.1 Terra cost?
Standard API rates are $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. OpenAI says standard pricing is 50% lower than for GPT-5.6 Terra and GPT-6.1 Sol. Token prices are only part of a deployment budget: the model documentation also lists cache-write charges, possible tool fees and higher rates for very long prompts. [1] [3]
The API documentation lists a 1,050,000-token context window and a maximum of 128,000 output tokens. It offers six reasoning-effort settings, with medium as the default. Developers who want tool calling must use the Responses API, according to OpenAI; the model also supports structured outputs and function calling there. [3]
In ChatGPT Work and Codex, OpenAI says the same plan allowance covers more tasks with Terra than with its predecessor and substantially more than with GPT-6.1 Sol or GPT-6 Astra. The announcement does not provide a universal number of tasks, so users should check their own plan and workspace limits. [1] [4]
What do the tests show?
On Agents’ Last Exam, a test of professional computer tasks, OpenAI reports that Terra at medium reasoning scored 15 percentage points above GPT-5.6 Terra at about one-third the cost. It also says Terra performed on par with GPT-6 Astra on that test at one-tenth the price per token. These are company-reported results under specific evaluation settings, not a promise for every application. [1]
For coding, OpenAI reports that Terra solved 50.3% of Terminal-Bench 4.0 tasks at maximum reasoning, compared with a published 21.5% for GPT-5.6 Terra. The company still recommends GPT-6.1 Sol or GPT-6 Astra for the most complex coding work. In an internal test designed to induce factual errors, it reports lower error rates than its predecessor, although errors were not eliminated. [1]
OpenAI’s system card says the company treats Terra as Critical in cybersecurity and High for biological and chemical capability under its Preparedness Framework, applying the safeguards stack used for Sol and Astra with model-specific updates. It cautions that its internal Codex deployment simulation is an additional risk signal, not a direct measure of external deployment safety. Teams considering automated actions should test their own workflows and retain appropriate human oversight. [5]
Reporting sources
- OpenAI · Introducing GPT-6.1 Terra ↗2026-10-26
- OpenAI Developers · API changelog: October 26, 2026 ↗2026-10-26
- OpenAI Developers · GPT-6.1 Terra model documentation ↗2026-10-26
- ChatGPT Learn · October 26–30, 2026 ↗2026-10-26
- OpenAI Deployment Safety Hub · Addendum to GPT-6 Astra System Card: GPT-6.1 Terra ↗2026-10-26
This article follows our editorial standards. To report an error, see our corrections policy.