OpenAI launches GPT-6.1 Terra: Astra-level computer use at a tenth of the price

OpenAI on Monday evening launched GPT-6.1 Terra, a fast, cheaper model built to use the computer on the user's behalf: reading what is on screen, clicking through applications, filling in forms and writing text. According to OpenAI, it matches its top model GPT-6 Astra on the company's main test of computer work, at one-tenth of the price per token.

The announcement came on OpenAI's X account and blog at 18:06 UTC (19:06 Oslo time) on 26 October 2026. The model is rolling out now to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and developers can access it through the API as gpt-6.1-terra.

What Terra is meant to do

OpenAI describes Terra as a model for "the communication and computer-based tasks that fill the workday". The company's own examples are drafting proposals, interpreting technical drawings, filling out forms and editing videos. In its API documentation, OpenAI also mentions matching product photos to catalogue entries and acting on application screens.

The model takes text and images as input and returns text. It has a 1,050,000-token context window, up to 128,000 output tokens and a knowledge cutoff of 1 June 2026. Audio and video are not supported. Computer use, file search, web search, code execution and MCP tools are available through the Responses API.

OpenAI is also clear about the limits: for demanding coding, research and deep reasoning, GPT-6.1 Sol and GPT-6 Astra remain its most capable models.

The numbers OpenAI presents

All results below are OpenAI's own measurements. The company has not published independent tests in the announcement.

• On Agents' Last Exam, a test of professional tasks in complex software, Terra scores 15 percentage points higher than its predecessor GPT-5.6 Terra at medium reasoning effort, at about one-third of the cost. OpenAI says Terra performs on par with GPT-6 Astra and completes tasks more than three times faster than both Astra and GPT-6.1 Sol.
• OpenAI also compares it with rival Anthropic: at max effort, Terra is said to be over three times as token-efficient as Claude Opus 5.5 at one-fourth of the price per token, which the company translates into roughly nine times lower cost per task.
• On the coding test Terminal-Bench 4.0, Terra solves 50.3% of tasks at maximum effort, against 21.5% for GPT-5.6 Terra.
• In OpenAI's internal test of error-inducing prompts, the error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7%. At maximum effort, the error rate is 4.3%.

Price and safety

Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. According to OpenAI, that is half the standard price of both GPT-5.6 Terra and GPT-6.1 Sol. Batch and Flex processing are 50% cheaper, while requests with more than 272,000 input tokens cost twice as much for input and 1.5 times as much for output. Regional processing adds a 10% premium where available.

In ChatGPT Work and Codex, the same plan is meant to cover more tasks with Terra than with GPT-5.6 Terra, and "substantially more" than with Sol or Astra. OpenAI does not give specific limits.

OpenAI writes that Terra meets the "Critical" threshold for cybersecurity under its Preparedness Framework, and that it keeps the protections developed for GPT-6 Astra. The company also says Terra showed strong resistance to prompt injection, meaning hidden instructions that try to divert the model from the user's task. Details are in the model's system card.

What leaders should note

This is hogby.ai's assessment, not OpenAI's claims.

• Automating screen work gets cheaper. OpenAI itself uses entering incoming invoices into accounting software as an example. The price makes it more realistic to test agents on high-volume routine work, but the gains must be measured on your own processes, not OpenAI's benchmarks.
• An agent that clicks needs access control. When the model acts in business systems on an employee's behalf, permissions, logging and approval of changes matter more than the choice of model.
• Pick the model per task. OpenAI itself recommends Sol or Astra for the heaviest coding and analysis. Many organisations will cut costs by using Terra for volume work and more expensive models where mistakes cost the most.

Sources and media

• OpenAI: Introducing GPT-6.1 Terra
• OpenAI on X: GPT-6.1 Terra announcement
• OpenAI on X: availability in ChatGPT Work, Codex and the API
• OpenAI Developers: GPT-6.1 Terra model documentation and pricing

Source: OpenAI. The sources were published on 26 October 2026 at 19:06 Oslo time and read by hogby.ai the same evening.

Illustration: AI-generated with Grok Bot / hogby.ai