GPT-6.1 Terra
GPT-6.1 Terra is OpenAI's newest model, released on October 26, 2026. It sits between Luna and Sol, is strong at reading images and writing naturally, and in our chat it thinks briefly before it answers.
gpt-6.1-terraContext 1.1M
GPT-6.1 Terra at a glance
GPT-6.1 Terra is OpenAI's mid-tier model, released on October 26, 2026. It is the newest GPT model as of November 2026. OpenAI's documentation describes it as a fast, efficient model for visual reasoning, precise computer use and clear, natural writing. It replaced GPT-5.6 Terra and costs less than its predecessor.
In our chat, GPT-6.1 Terra is the model behind the Think chip. It reasons briefly before answering, which makes it noticeably better than GPT-6 Luna at maths, logic and careful writing. Chat with GPT-6.1 Terra for free in your browser.
- Maker: OpenAI (GPT.me is an independent site, not run by OpenAI).
- Reads text and pictures; writes text only.
- Context window: 1,050,000 tokens on OpenAI's platform, up to 128,000 output tokens.
- Knowledge cut-off: June 1, 2026.
- API price: $1 per million input tokens, $0.05 for cached input and $5 per million output tokens (OpenAI list price, Nov 2026).
What "6.1" and "Terra" mean
OpenAI's naming has two parts. The number is the generation: 6.1 is an update within GPT-6, released a few weeks after the first GPT-6 models. The word is the tier. Terra sits in the middle, between the inexpensive Luna and the high-end Sol, with Astra above them all as the flagship.
In ChatGPT, OpenAI made GPT-6.1 Terra available in the Work and Codex areas for paid plans, including Plus, Pro, Enterprise and Edu. The free ChatGPT chat runs GPT-6 Luna instead. On GPT.me you can use Terra without a subscription, within the same daily allowance as Luna.
OpenAI is open about Terra's ceiling: in its announcement it says GPT-6.1 Sol and GPT-6 Astra remain the better choice for the most complex coding tasks. Terra aims at strong everyday quality at a moderate price, not at the top of every chart.
What Terra does better than Luna
Thinking before answering pays off whenever a question has several steps or a trap in it:
- Maths and word problems. It works through the steps instead of jumping to a number. Try a discount puzzle, or use the math solver.
- Logic and planning. Schedules with constraints, comparing options against several criteria, spotting the flaw in an argument.
- Images that need thought. Reading a chart and drawing a conclusion, working out what a diagram shows, checking a handwritten calculation in a photo.
- Natural writing. OpenAI highlights clear writing as one of Terra's strengths. Use it for essays, speeches and messages where tone matters. A wedding toast.
- Code. Finding the bug in a function, explaining an error message, writing tests. The code assistant is set up for this.
How Terra works in our chat
Press Think in the message box and the chip lights up. Your next message goes to GPT-6.1 Terra with a low reasoning setting: enough to plan the answer and check it, while still replying within seconds. While it thinks you see a short indicator, and afterwards a "Thought for N s" label shows how long it took. The reasoning itself stays hidden; you get the finished answer.
Terra may write longer answers than Luna in our chat, which helps for essays and step-by-step solutions. Like Luna, it reads one attached image per message, and it shares the same daily allowance. The limits of our site apply too: no web search, no voice, no picture creation, no memory between chats, and long conversations are trimmed so the model only sees the recent part.
Press Think again to go back to Luna for quick questions. There is no need to keep Terra on all the time; Luna is faster for simple requests.
A practical pattern: draft with Luna, then switch on Think and ask Terra to review the result. For example, let Luna write a first version of a cover letter, then ask Terra to point out weak sentences and anything that sounds generic. The same works for homework solutions, where Terra can check each step, and for code, where it can look for cases the first version did not handle. The homework helper and essay writer pages are good places to try this.
Limits to keep in mind
- It can still be wrong. Reasoning reduces careless slips but does not remove them. Check facts, figures and citations that matter. Our guide to fact-checking AI answers helps.
- Knowledge ends in mid-2026. Anything after June 1, 2026 is unknown to it unless you paste it in.
- Not the top model. For very hard research or large codebases, OpenAI's own guidance points to GPT-6.1 Sol or GPT-6 Astra, which are not in our web chat.
- Slower than Luna. Short questions take a moment longer because of the thinking step.
Terra for developers
Through the API the model id is gpt-6.1-terra on OpenAI and openai/gpt-6.1-terra on OpenRouter. It supports reasoning effort from none to max. OpenAI's documentation notes that tool calling with this model requires the Responses API. Our GPT API guide has examples, and reasoning models explained covers what the effort setting changes.
FAQ
Is GPT-6.1 Terra free on GPT.me?
Yes. Press the Think chip in the chat. Terra uses the same daily allowance as Luna, with a few free messages every day and no account.
Is GPT-6.1 Terra better than GPT-6 Sol?
No. Sol is a higher tier and stronger on complex coding and analysis. Terra is cheaper and faster, and good enough for most everyday tasks that need some reasoning.
Why does Terra take longer to answer?
It reasons briefly before writing. That short pause is what makes it better at multi-step questions. The "Thought for N s" label shows how long it took.
Can I see Terra's reasoning?
No. In our chat the reasoning stays hidden and you get the final answer. You can ask it to show the steps of its solution in the answer itself.
Is GPT-6.1 Terra in ChatGPT?
Yes, in ChatGPT's Work and Codex areas for paid plans such as Plus and Pro. The free ChatGPT chat uses GPT-6 Luna.
Got a question? Ask it now
No account and no download. Open the chat, type what is on your mind, and the answer starts appearing in a second or two.
Start chatting