Tired of copy-paste chores? GPT-6.1 Terra does the clicking for less

👋 Easy read · Oct 27, 2026 · ~7 min read

The short version

OpenAI’s new Terra model takes on your screen chores for less.

GPT-6.1 Terra, out Oct 26, 2026, reads what’s on your screen, clicks and types across your apps, and writes in a clear, direct style.

On paid ChatGPT plans it lives in ChatGPT Work and Codex, where OpenAI says your usage allowance covers more tasks with it than with its bigger models. Early users are already filling in forms, making fonts and hunting bugs with it, and a few are unimpressed.

Jargon jar: a model is the AI brain behind an app like ChatGPT, and paid plans let you pick between several. Computer use means the AI looks at your screen and moves the pointer, clicks and types, the way you would.

ChatGPT Work is the part of ChatGPT built for longer, multi-step jobs, next to the regular Chat; Codex is OpenAI’s coding helper. Your usage allowance is how much AI work your plan includes in a stretch of time.

Reasoning effort is how long the model thinks before it acts: higher is slower and uses more of your allowance. The API is how developers build OpenAI’s models into their own apps.

Why people are talking about it

OpenAI announced GPT-6.1 Terra on Oct 26, 2026, calling it a fast, efficient model for “detailed visual reasoning and precise computer use” (OpenAI: introducing GPT-6.1 Terra).

Its launch post on X passed 25,000 likes within a day (OpenAI on X: GPT-6.1 Terra launch post), and the Hacker News thread passed 600 points and 380 comments (Hacker News discussion: GPT-6.1 Terra, numbers as of our research).

The busywork problem

Think of the jobs that are mostly your eyes and your mouse: copying invoice details into accounting software, checking product photos against a catalog, filling in a long web form from a PDF. None of it is hard, but it eats the afternoon.

OpenAI says Terra is built for exactly “the communication and computer-based tasks that fill the workday”, from drafting proposals and reading technical drawings to filling out forms and editing videos (OpenAI: introducing GPT-6.1 Terra).

The second pain is quieter: big models can drain your plan fast. Terra goes after both.

What Terra makes possible

It looks first, then acts. Terra reasons about images and screens, then takes action across your browser, desktop and apps (ChatGPT Learn: what’s new, Oct 26–30, 2026).

It writes like a helpful colleague. Per the same page, it puts results first and explains the details simply.

Your plan stretches further. OpenAI says the same usage allowance covers more tasks with Terra than with GPT-5.6 Terra, and substantially more than with GPT-6.1 Sol or GPT-6 Astra.

It also says Terra finishes tasks over three times faster than both of those bigger models (OpenAI: introducing GPT-6.1 Terra).

Your plan’s overall usage limits don’t change, and some Plus and Pro users may now see Terra at Medium in Work’s default Power presets, in place of GPT-6.1 Sol at Light (OpenAI Help Center: ChatGPT release notes).

“Use Terra to reason about images and screens, then take action across your browser, desktop, and apps.” ChatGPT Learn, OpenAI’s guide to ChatGPT Work and Codex

OpenAI’s guides name specific jobs. Here is how they line up with everyday headaches (ChatGPT Learn: what’s new, Oct 26–30, 2026; ChatGPT Learn: models in Work and Codex):

The headache	What Terra can do	What stays with you
A supplier lookbook and a messy catalog	Match product photos to catalog entries and update the listings	Checking each match before it goes live
Campaign images going out today	Check them for cropping and text placement, then organize the edits	Signing off on the final images
A wall of dashboard screenshots	Summarize the main changes, citing chart labels and values	Checking anything it says it can’t read confidently
Forms to fill from paperwork	Fill out forms from source documents	Comparing names and numbers with the originals
A proposal due Friday	Draft it in a results-first style	Your edits and your send button

What early users did on day one

These are public posts on X from the first 24 hours after launch, one person or team each: their own reports, not independent reviews.

The form race. Victor E. Nunez, who works on ChatGPT and Codex at OpenAI per his X profile, had Terra and then GPT-6.1 Sol fill out the same long web form in Fast mode. In his run, Terra took 53.5 seconds and Sol took 3 minutes 21.7 seconds (Example shared by @victornunez). One run, by OpenAI staff.
A font from handwriting. Developer Chirag Patil turned their handwriting into a font with Terra in ChatGPT Work, to use in blog posts (Example shared by @VectorQl).
From photo to 3D. The team behind the ALwithU agent app gave Terra a seaside photo and asked it to recreate the coffee mug in Blender, a 3D design app. Terra worked directly in Blender through computer use (Example shared by @ALwith_ai).
A bug hunt. Developer ArcyDev pitted Terra against GPT-6.1 Sol on real bugs from a game project. Terra at its highest reasoning setting found 3 more bugs that were still in the code, finished faster and cost less by their estimate (Example shared by @ArcyDev).

Not everyone is cheering

Web design fell flat for one tester. Mia, who runs a build-100-web-pages challenge on many models, called Terra’s pages repetitive and said it missed basic instructions, advising others to skip it for web design (Example shared by @MiaAI_lab).
Usage may not feel lighter yet. Software engineer Trí ACJ found Terra really fast, but said it seemed to eat a lot of their 5-hour quota even with lower token use (Example shared by @TriACJ1409).

OpenAI is upfront about the ceiling too: for the most complex coding it still points to GPT-6.1 Sol and GPT-6 Astra, and for hard screen jobs that need visual judgment its guide suggests GPT-6 Astra (ChatGPT Learn: computer use).

Where it stops

Not in regular Chat. Terra is a model for ChatGPT Work and Codex on Plus, Pro, Business, Enterprise and Edu plans, and it is still rolling out (OpenAI Help Center: ChatGPT release notes).
Computer use needs setup. It runs in the ChatGPT desktop app on Mac or Windows, in supported regions, through a Computer Use plugin. On a Mac you grant Screen Recording and Accessibility permissions; on Windows it takes over your pointer and keyboard while it works (ChatGPT Learn: computer use).
It asks before it touches an app. ChatGPT asks permission before using each app, unless you chose Always allow, and may ask again before sensitive or disruptive actions.
Keep secrets out of it. OpenAI’s guidance: one clear app at a time, close sensitive apps, and avoid tasks that need passwords or other secrets unless you are there to approve each step.

How to try it

In the ChatGPT desktop app or ChatGPT Work on the web, open Advanced under the message box and choose GPT-6.1 Terra from the model list. Work on mobile supports Terra too (ChatGPT Learn: what’s new, Oct 26–30, 2026).
For screen chores, open Plugins > Computer Use in the desktop app, install the plugin, switch on its toggles and select Try now (ChatGPT Learn: computer use).
Coding in the terminal? Run codex -m gpt-6.1-terra on Codex CLI 0.159.0 or later (ChatGPT Learn: models in Work and Codex).
Building your own app? Developers get gpt-6.1-terra in the API at $1 per million input tokens and $5 per million output tokens, billed apart from ChatGPT plans (OpenAI Help Center: model release notes).

Chef’s pick: pick one weekly chore that is mostly copying from one screen to another, give Terra one clear app to work in, and check every name and number against the original before anything gets sent.

Sources

Show all 14 sourcesShow fewer