OpenAI introduced GPT-6.1 Terra on October 26, 2026, a fast, lower-cost model built for the work that fills an ordinary day: reading screens and images, filling out forms, and writing clearly. It is available now in ChatGPT Work, Codex, and the API as gpt-6.1-terra.

Coffee summary

  • FACT: OpenAI says Terra's standard API prices are 50% lower than GPT-5.6 Terra and GPT-6.1 Sol.
  • FACT: On OpenAI's internal benchmark of error-inducing prompts, the error rate at low reasoning effort fell from 53.6% (GPT-5.6 Terra) to 16.7%.
  • FACT: On Terminal-Bench 4.0 at max effort, Terra solves 50.3% of tasks versus 21.5% for GPT-5.6 Terra.
  • FACT: In ChatGPT it is available in Work and Codex on Plus, Pro, Business, Enterprise, and Edu plans, not in regular Chat.
  • OPINION: Terra is the model to test first for high-volume, screen-based office work; keep Sol or Astra for the hardest coding.

What Terra is for

OpenAI pitches Terra as a model that sees and acts. Examples from the launch include turning architectural drawings into a 3D model, reviewing an animated film shot by shot to build a time-coded log, and reading invoices to enter their details into accounting software. ChatGPT's own help notes add practical ideas, such as matching supplier product photos to catalog entries or checking campaign assets for cropping and text placement.

The numbers OpenAI is leading with

On Agents' Last Exam, a set of professional computer tasks, OpenAI says Terra at medium effort scores 15 percentage points higher than GPT-5.6 Terra at about one-third of the cost, and performs on par with GPT-6 Astra at one-tenth its price per token. OpenAI also says Terra finishes tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.

CLAIM: comparisons with competitors, such as being about 9x cheaper per task than Claude Opus 5.5 at max effort, come from OpenAI's own testing. Useful as a signal, but run your own workload before you switch.

OpenAI is also honest about limits: for the most complex coding tasks, it says GPT-6.1 Sol and GPT-6 Astra remain the better choice.

Fewer made-up answers

The factuality gains may matter most for everyday users. At maximum effort, OpenAI reports the error rate on its hard-prompt benchmark dropped to 4.3%. That is a big improvement, but not zero, so anything that goes to a customer or into your books still needs a human check.

How to try it

In the ChatGPT desktop app or ChatGPT Work on the web, choose GPT-6.1 Terra from the model list under Advanced. Browser use and computer use each have their own setup and access requirements, and enterprise access depends on what your administrator has turned on.

What this means for you

  • Small businesses: test Terra on one repetitive screen task, like invoice entry or listing updates, and measure time saved.
  • Developers: swap Terra into high-volume pipelines where cost matters and accuracy can be spot-checked.
  • Writers and marketers: OpenAI says Terra writes in a direct, results-first style; compare it against your current drafts.

Bottom line

GPT-6.1 Terra makes capable computer use cheaper and faster, according to OpenAI's numbers. Start with one boring, repetitive task, keep a human in the loop, and let your own results decide.

*Sources: OpenAI, "Introducing GPT-6.1 Terra"; ChatGPT Learn, What's new Oct 26–30, 2026.*