OpenAI launches GPT-6.1 Terra: What the new ChatGPT model means for users and developers

What to know

• Who gets it: Plus, Pro, Enterprise and Edu subscribers, in ChatGPT Work and Codex, according to OpenAI's announcement. OpenAI's ChatGPT release notes also list Business plans. Access depends on plan, region and workspace settings as the rollout reaches each account.
• Not in regular chat: OpenAI's help pages say GPT-6.1 Terra is a model option for Work and Codex and is not available in standard ChatGPT conversations.
• Free users: OpenAI hasn't listed the Free or Go plans among those getting GPT-6.1 Terra. Whether or when free users will get it is unknown.
• Developer price: $1 per million input tokens, 5 cents per million cached input tokens and $5 per million output tokens, according to OpenAI's developer documentation.
• Context window: about 1.05 million tokens, with up to 128,000 output tokens.
• Performance claims: OpenAI's benchmark results, including its comparisons with Anthropic's Claude, come from the company itself and have not been independently tested.

OpenAI on Monday, Oct. 26, released GPT-6.1 Terra, a new version of its mid-priced AI model that the company says is faster, cheaper and better at operating a computer on a user's behalf than the model it replaces. Those performance claims come from OpenAI and have not been independently tested.

The model is rolling out to paying ChatGPT subscribers inside ChatGPT Work and Codex, OpenAI's workspace and coding tools, and is available to developers through the company's API under the name gpt-6.1-terra. It succeeds GPT-5.6 Terra. Standard API prices are $1 per million input tokens and $5 per million output tokens, which OpenAI says is half the price of both GPT-5.6 Terra and its more powerful GPT-6.1 Sol model.

The launch landed on the same day rival Anthropic filed publicly for an initial public offering, a reminder of how quickly the largest AI companies are competing on price as well as capability.

Who can use GPT-6.1 Terra, and where

For most people, the key detail is where the model shows up. OpenAI sorts its current lineup into four classes, from the least to the most capable: Luna, Terra, Sol and Astra, as 9to5Mac noted in its coverage. Terra sits in the middle, meant to balance capability, speed and cost.

According to OpenAI's help center, GPT-6.1 Terra is offered inside ChatGPT Work and Codex, not in ordinary back-and-forth ChatGPT chats. Eligible subscribers can pick it from the model list once the rollout reaches their account. The release notes say plan usage limits are unchanged, and that some Plus and Pro users may see GPT-6.1 Terra set as a default in certain Work presets in place of GPT-6.1 Sol.

OpenAI says the practical effect for subscribers is that the same usage allowance stretches further, covering more tasks than with GPT-5.6 Terra and substantially more than with its larger Sol and Astra models.

Free ChatGPT users are not named in the announcement or the release notes. OpenAI's pricing page says free accounts get limited access to Codex and to ChatGPT Work on desktop, but the company hasn't said whether GPT-6.1 Terra will be part of that. NewsCubby could not confirm any timeline for free users.

What OpenAI says is different

OpenAI describes GPT-6.1 Terra as built for everyday office work: drafting documents, reading technical drawings, filling out forms and working inside desktop apps. The company says it also writes more clearly and naturally than its predecessor.

The model accepts text and images as input and produces text, according to the developer documentation. It does not support audio or video input. Its knowledge cutoff is June 1, 2026.

The headline feature is what the industry calls computer use, in which an AI model looks at a screen and carries out steps such as clicking, typing and moving between programs. OpenAI's example is a business that wants software to read incoming invoices and type the details into accounting software, a high-volume task where cost per job matters.

OpenAI's own performance claims

The following figures come from OpenAI's announcement. They have not been independently tested, and benchmark results don't always reflect how a model performs in real use.

• Computer use: On a benchmark called Agents' Last Exam, OpenAI says GPT-6.1 Terra scores 15 percentage points higher than GPT-5.6 Terra at medium reasoning effort, at about a third of the cost.
• Cost versus Anthropic: At maximum effort, OpenAI says the model costs about one-ninth as much per task as Anthropic's Claude Opus 5.5, based on its own calculations of token use and price. NewsCubby found no public response from Anthropic to that comparison as of Monday afternoon.
• Fewer errors: On an internal OpenAI test built to provoke mistakes, the company says the factual error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7%. At maximum effort, OpenAI says the error rate was 4.3%.
• Coding: OpenAI says the model solved 50.3% of tasks on the Terminal-Bench 4.0 coding test at maximum effort, more than double GPT-5.6 Terra's published score. The company also says its Sol and Astra models remain the better choice for the hardest coding work.

Even by OpenAI's own figures, the model still makes factual errors, so users should keep checking important answers.

What it costs developers

For businesses and developers that pay by usage, the price change is the most concrete part of the announcement. OpenAI's documentation lists standard prices of $1 per million input tokens and $5 per million output tokens. A token is a small chunk of text, roughly three-quarters of an English word on average.

The documentation includes some fine print:

• Requests with more than 272,000 input tokens are billed at higher rates, double for input and 1.5 times for output, across the whole request.
• Batch and Flex processing cost half the standard price, while the faster "fast mode" costs double.
• Reused, or cached, input costs 5 cents per million tokens.

Before launch, the tech site TestingCatalog reported on Oct. 20 that the GPT-6.1 Terra name had appeared in evaluation settings in a Microsoft code repository, citing a post by an independent AI benchmarking account. The site also pointed to a paywalled report from The Information that OpenAI was preparing a new Terra model to compete with Anthropic on price. OpenAI didn't comment on those reports before Monday's launch.

Safety ratings

OpenAI published a safety document alongside the model. Under its internal Preparedness Framework, the company says it treats GPT-6.1 Terra as reaching its "Critical" level for cybersecurity capability and "High" for biological and chemical capability, and that it uses the same set of safeguards as its larger GPT-6 models. Those ratings describe what OpenAI believes the model could be capable of, and determine which protections the company applies, according to the document.

OpenAI also says that in its own testing the model was more likely than its predecessor to stay on task and to tell users about its limitations, and that it resisted attempts to hijack it with hidden instructions, known as prompt injection. Those results also come from OpenAI and have not been independently tested.

The AI race context

Monday's release came as OpenAI's main rivals were making news of their own. Anthropic, the maker of Claude, filed publicly for an IPO on the same day. NewsCubby covered that filing earlier today: Anthropic IPO: What the S-1 Filing Reveals for Investors.

TestingCatalog's pre-launch report also cited The Information as saying Google was preparing a new Gemini model and Anthropic a new Claude model this month, though those plans weren't confirmed by the companies and some releases could slip. Google hasn't announced a response to GPT-6.1 Terra.

OpenAI's decision to compare its new mid-tier model directly with Anthropic's top model on cost shows where the competition is heading. Businesses that run AI on millions of tasks pay close attention to price per job, and companies are increasingly marketing cheaper models that, they say, come close to their flagships.

What this means for you

• If you pay for ChatGPT Plus or Pro: Look for GPT-6.1 Terra in the model list in ChatGPT Work or Codex. It may not appear right away while the rollout continues.
• If you use free ChatGPT: Nothing changes in your regular chats today, and OpenAI hasn't said when or whether free users will get the model.
• If you're a developer: The model is available now in the API as gpt-6.1-terra. Check the long-prompt pricing before moving large workloads.
• For everyone: Treat the benchmark figures as the company's own claims until independent testers weigh in.

*Credits: reporting from OpenAI, The Verge, 9to5Mac and TestingCatalog. Photo: HaeB / Wikimedia Commons, CC BY-SA 4.0 (archive photo). NewsCubby articles are produced with AI assistance and reviewed by humans.*

*Want more? Explore today's headlines on NewsCubby and join My Day free to make NewsCubby part of your day.*