ChatGPT's new model GPT-6.1 Terra: what changes for your business
A new model is appearing in paid ChatGPT plans. OpenAI says it makes far fewer factual mistakes than the one it replaces and costs developers half as much. Here is what that means for a small business, and the one test worth running this week.
On this page
OpenAI has added a new model, GPT-6.1 Terra, to ChatGPT's paid plans. If you or your team use ChatGPT for office work, a new name is appearing in your model list, and OpenAI says it makes far fewer factual mistakes than the model it replaces and costs developers half as much. This update explains what is new, which plans get it, and what to check in your account this week. It is based on OpenAI's own announcement and the ChatGPT release notes, read on 28 October 2026.
What happened: OpenAI released GPT-6.1 Terra for paid ChatGPT plans
- OpenAI's announcement of 26 October 2026 describes GPT-6.1 Terra as "a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing".
- The same announcement lists the work it is built for: drafting proposals, reading technical drawings, filling out forms and editing videos. For developers it gives the example of reading incoming invoices and entering their details into accounting software.
- The ChatGPT release notes say the rollout covers the Plus, Pro, Business, Enterprise and Edu plans, inside the Work and Codex experiences, and that it reaches accounts gradually. Your plan's overall usage limits are unchanged, but the notes add that some Plus and Pro users may see a different default: GPT-6.1 Terra at Medium instead of GPT-6.1 Sol at Light.
- OpenAI says the standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol, at 1 dollar per million input tokens and 5 dollars per million output tokens. OpenAI does not publish a rupee price for the API, so these are the dollar prices as its page shows them.
- On OpenAI's own internal set of error-inducing prompts, the error rate at low reasoning effort fell from 53.6% with the earlier GPT-5.6 Terra to 16.7% with GPT-6.1 Terra, and to 4.3% at maximum effort.
- OpenAI says that in its own tests the model resisted hidden instructions planted to pull it away from the user's task, which matters when staff paste outside documents into it.
- What is not known: neither page mentions the Free or Go plans, so there is no date for them, there is no rupee price for the API, the error-rate figures are OpenAI's own and untested by outsiders, and neither page says anything about Hindi or regional language quality.
Why it matters for a small business
Two things here touch the money side of your business.
The first is wrong answers. The biggest cost of AI in a small business is not the subscription, it is the hour you lose when a confident wrong answer reaches a customer. OpenAI claims a large drop in that kind of error on its own test set, which is a reason to retest the tasks you gave up on earlier this year, not a reason to stop checking output.
The second is the work it is aimed at. Reading a form, a bill or a drawing and then acting on it is the slow middle of most small businesses. OpenAI has priced that work lower for developers, which usually means the tools you buy from software vendors get cheaper or faster over the coming months.
Who should care
- Owners and teams on a paid ChatGPT plan, who will see the new model in their list.
- Anyone who uses ChatGPT to read documents: bills, quotations, delivery challans, site drawings, handwritten order slips.
- Owners who pay an agency or developer to build on OpenAI's API, where the price change lands on your bill.
- Owners on free ChatGPT or on ChatGPT Go, where this model is not in the announced list.
What changes for you
| What | Before | After this release |
|---|---|---|
| Model list in ChatGPT | Earlier GPT-6 models only | GPT-6.1 Terra also appears, on paid plans |
| Plans that get it | Not released yet | Plus, Pro, Business, Enterprise and Edu, gradually |
| Usage limits on your plan | As before | Unchanged, and OpenAI says the same allowance covers more tasks |
| API price for developers | Earlier price | 50% lower, 1 dollar in and 5 dollars out per million tokens |
| Wrong answers | 53.6% error rate at low effort on OpenAI's test | 16.7% at low effort, 4.3% at maximum effort |
Our own advice: treat the error-rate drop as a promise to verify, not a result. Keep the person who checks the output in place, especially for anything that reaches a customer or an auditor.
What to do this week
- Open ChatGPT, click the model list, and see whether GPT-6.1 Terra is there. If it is not, your account has not received it yet and there is nothing to install.
- Take one document task you do every week, for example reading five supplier bills into a sheet. Run it once on your old model and once on GPT-6.1 Terra, with the same instruction.
- Check both outputs against the actual papers, line by line, and count the mistakes. That number, not OpenAI's, decides whether you change how you work.
- If the new model wins, write the working instruction down for your team. Our guide on how to write a good prompt for business shows the shape that keeps results steady.
- If staff now start feeding customer papers into ChatGPT, fix the ground rules first: simple AI rules for your team.
- If you pay a vendor for an OpenAI-based tool, ask in writing which model they use now and whether the lower price reaches you. Before you renew, what to check before you pay for an AI tool is a short checklist.
What to watch next
- Whether OpenAI brings GPT-6.1 Terra to the Free and Go plans, which many first-time users in India are on.
- Whether independent testers confirm the drop in factual errors, or find it holds only on OpenAI's own prompts.
- Your own vendor bills over the next quarter, to see whether the API price cut actually reaches your invoice.
Not sure where AI can help your business safely? The free AI Business Checkup takes about three minutes.
