On October 26, 2026, OpenAI announced GPT-6.1 Terra, a new model that it describes as fast and efficient. I use ChatGPT for my store, so I read the announcement. I have not tested Terra myself, so this is not a review. It is a summary of what OpenAI says, in my own words, plus what I would check before believing it. I read the Spanish version of the page and translated the main points myself, with Claude’s help.
The short version
OpenAI presents Terra as a cheaper and faster model that is good at working with a computer (filling in forms, editing video, reading technical drawings), at writing more naturally, and at making fewer factual mistakes. It is not the company’s most powerful model: for the hardest coding tasks, OpenAI itself says GPT-6.1 Sol and GPT-6 Astra are still the better choice.
Who can use it, according to OpenAI
| You are… | What OpenAI says |
|---|---|
| A ChatGPT user | Available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, from the day of the announcement. The page does not mention a free plan. |
| A developer | Available in the OpenAI API as gpt-6.1-terra. |
OpenAI also says that, with the same usage limit, your plan lets you do more tasks with Terra than with GPT-5.6 Terra, and many more than with GPT-6.1 Sol or GPT-6 Astra.
Price for developers
- $1 per million input tokens.
- $0.05 per million cached input tokens.
- $5 per million output tokens.
OpenAI says these standard prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol. Its example of what that allows: automating large volumes of work, such as reading incoming invoices and entering the data into accounting software.
The main claims
- Computer work: on a test of professional computer tasks called Agents’ Last Exam, OpenAI says Terra at medium effort scores 15 percentage points higher than GPT-5.6 Terra, at about a third of the cost. It also says Terra completes tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.
- Fewer factual errors: on an internal test with misleading prompts, OpenAI says the error rate at low effort went from 53.6% with GPT-5.6 Terra to 16.7% with Terra, and from about 22% to 4.3% at maximum effort. The 22% is my own calculation from OpenAI’s “18 points lower” figure.
- Coding: on Terminal-Bench 4.0 at maximum effort, OpenAI says Terra solves 50.3% of the tasks, against 21.5% for GPT-5.6 Terra.
- Comparison with Claude: OpenAI says that at maximum effort Terra uses more than three times fewer tokens than Claude Opus 5.5, with a price per token a quarter as high, which it says makes the cost per task about nine times lower. This is OpenAI’s comparison with a competitor, so I would want to see an independent one.
- Safety: OpenAI says Terra reaches the cybersecurity threshold that its Preparedness Framework calls “critical” and keeps the protections built for GPT-6 Astra. It also says Terra is more likely than the previous version to stay within the task and to tell you honestly about its limits, and that it resisted prompt injection well in its tests.
What I find interesting for everyday use
The claim that matters most to me is the one about fewer factual errors. If a model makes things up less, the habit of checking everything is still necessary, but it hurts less. I tested this kind of behavior with ChatGPT in my test of AI references, where checking made the difference between real and invented sources.
What I want to test
- Does Terra make up fewer references than what I found before?
- Does it respect word limits better? I measured that in my word-limit test.
- Does “clearer and more natural writing” mean less repetition, like the one I found in my repetition test?
- Is it noticeably faster in the app, not only on OpenAI’s benchmarks?
What to keep in mind
- All the numbers above come from OpenAI’s own announcement and its own tests. I have not verified any of them.
- The error-rate result comes from an internal benchmark that OpenAI built, with prompts designed to mislead the model. It shows a trend, not what you will see in your own questions.
- Model names, plans and prices change. If something here no longer matches OpenAI’s page, trust the page.
Source
- OpenAI, “Introducing GPT-6.1 Terra”, October 26, 2026. I read it on November 8, 2026.

