OpenAI announced GPT-6.1 Terra on October 26, 2026, and the model reached paying ChatGPT users the next day. It is the newest member of the company’s Terra line, the mid-range tier built for everyday work that happens on a computer screen rather than in a chat box.
GPT-6.1 Terra reads what is on screen, clicks through applications, fills in forms, and drafts documents. It combines visual reasoning with computer use, so it can act inside software instead of only answering questions about it. OpenAI presents it as a faster and cheaper way to automate repetitive office work, not as a replacement for its most capable models.
Price is the headline. OpenAI says standard API rates for GPT-6.1 Terra are 50 percent lower than those of GPT-5.6 Terra and GPT-6.1 Sol, at $1 per million input tokens and $5 per million output tokens in USD. Every performance claim behind that price comes from OpenAI’s own evaluations, and this article labels those claims as the company’s throughout.
The Short Version: GPT-6.1 Terra in Brief
- Announced October 26, 2026, and available from October 27, 2026 to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex. Developers can call it through the OpenAI API as gpt-6.1-terra.
- Built for visual reasoning and computer use: reading screens, acting across apps, filling out forms, drafting proposals, interpreting technical drawings, and editing videos.
- Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens in USD.
- OpenAI reports 15 percentage points higher than GPT-5.6 Terra on its Agents’ Last Exam test at medium effort, and 50.3 percent on Terminal-Bench 4.0 at maximum effort, against 21.5 percent for its predecessor. These are the company’s own figures.
- OpenAI still points users to GPT-6.1 Sol and GPT-6 Astra for the most complex coding and in-depth research. The model runs only on OpenAI’s servers, with no published weights.
How GPT-6.1 Terra Uses a Computer
For GPT-6.1 Terra, computer use means AI that operates software the way a person does. The model looks at the screen, finds the right controls, clicks buttons, enters text, and moves between applications to finish a multi-step job. That demands more than answering questions: it must understand a visual interface, remember what it has done, and recover when a screen looks unexpected.
OpenAI describes the model this way: “We’re introducing GPT-6.1 Terra, a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing.” The company calls it a significant step up from GPT-5.6 Terra for drafting proposals, interpreting technical drawings, filling out forms, and editing videos.
Chief executive Sam Altman announced the model on X on October 26, 2026, writing: “6.1 Terra is here! It is good at a lot of things, but we are especially excited for this new era of computer use.” The post signals OpenAI’s bet: agents that act inside software, priced for volume use.
The examples OpenAI gives for GPT-6.1 Terra are deliberately ordinary: reading incoming invoices into accounting software, matching product photos to catalog entries, investigating a codebase, and building agents across business systems. The model takes text and image input and returns text, with a 1.05-million-token context window, up to 128,000 output tokens, and a knowledge cutoff of June 1, 2026. Reasoning effort runs from none to maximum, with medium as the default.
GPT-6.1 Terra Pricing and Who Gets It
The 50 percent cut for GPT-6.1 Terra applies to standard API rates, the prices developers pay to call the model in their own applications. OpenAI lists $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens, all in USD, and says those rates are 50 percent lower than GPT-5.6 Terra and GPT-6.1 Sol and about one-tenth the per-token price of GPT-6 Astra.
Qualifications matter. Prompts above 272,000 input tokens use higher rates, with the whole request repriced.
Batch and Flex processing cost 50 percent less than standard, regional processing adds 10 percent where available, and Fast mode costs twice the standard rate. Teams automating long documents should test their own prompt sizes before assuming the headline rate.
Access follows the paid tiers. Plus, Pro, Enterprise, and Edu users get GPT-6.1 Terra in ChatGPT Work and Codex, and developers can call it as gpt-6.1-terra. It is also generally available on Amazon Bedrock.
Free users are not listed for this model and keep access to GPT-5.6 Luna. OpenAI says the same usage allowance covers more tasks than with GPT-5.6 Terra, and substantially more than with GPT-6.1 Sol or GPT-6 Astra.
Why You Should Care
The honest case for GPT-6.1 Terra is economic. Most office automation does not need the smartest model. It needs one good enough, fast enough, and cheap enough to run thousands of times a day.
A model that halves the token price and finishes tasks faster changes which workflows are worth automating: invoice intake, form processing, catalog matching, and document drafting.
Early independent evidence is narrower than the launch headline. Artificial Analysis scored the model at 48.1 on its Intelligence Index at maximum effort, up from 42.1 for GPT-5.6 Terra, at $0.40 per task instead of $1.40. The same analysis placed its low and medium effort settings on the cost-efficiency frontier, while broader tests put it below GPT-6.1 Sol and GPT-6 Astra.
Buyers should therefore test finished work, not token prices. Run the same workflow through GPT-6.1 Terra at low and medium effort and through GPT-6.1 Sol, then compare the finished work and the bill.
A cheaper model that needs more retries or more human review can cost more per finished task. OpenAI’s own charts show the per-task comparison depends on the effort setting and the test.
What OpenAI Has Not Proven
Every headline number for GPT-6.1 Terra comes from OpenAI’s own evaluations. No independent lab had reproduced the Agents’ Last Exam result, the Terminal-Bench 4.0 score, or the factual-error rates at the time of writing.
That does not make the figures false. It means readers should treat them as the vendor’s case, tested on the vendor’s chosen tasks.
The safety picture is mixed, by OpenAI’s own account. The launch material classifies the model at the Critical cybersecurity threshold under the company’s Preparedness Framework, with protections carried over from GPT-6 Astra. A system-card addendum summarized in launch coverage reports lower scores than GPT-5.6 Terra on one dynamic mental-health evaluation, and higher scores than GPT-6.1 Sol on some simulated unauthorized-action measures.
OpenAI also says the model resisted prompt-injection attempts designed to divert it. Teams deploying agents should run their own safety checks.
Practical questions remain. OpenAI has not published the model’s weights or parameter count, so it cannot run locally. Real-world error rates, long-prompt repricing effects, and regional availability were not fully spelled out.
None of that stops the model being useful. All of it argues for a trial on real work first.
Quick Answers
Who can use GPT-6.1 Terra today?
Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, from October 27, 2026. Developers can call it as gpt-6.1-terra, and it is also on Amazon Bedrock.
How much does it cost through the API?
Standard rates are $1 per million input tokens, $0.05 cached, and $5 output, all in USD. OpenAI says that is 50 percent lower than GPT-5.6 Terra and GPT-6.1 Sol, with higher rates for very long prompts.
Is GPT-6.1 Terra better than GPT-6.1 Sol?
It depends on the job. OpenAI reports that GPT-6.1 Terra beats its predecessor on agent benchmarks at lower cost for routine work, while Sol and Astra remain the better choices for the most complex coding and in-depth research.
Is it free?
No. The launch names paid plans only, and API use is billed per token. Free users keep access to GPT-5.6 Luna, according to launch coverage.
What should buyers check before deploying it?
Test it on real workflows at each effort setting, compare cost per finished task rather than cost per token, and run safety checks on prompt injection and task boundaries. The headline figures are OpenAI’s own evaluations.
The Takeaway on GPT-6.1 Terra
GPT-6.1 Terra is OpenAI’s pitch for the workday at scale: a model that reads screens, acts inside software, writes naturally, and costs half as much per token as the model it replaces. The verified facts are the launch date, the price list, the access tiers, and the company’s own benchmark claims, each labeled as such above.
For teams with repetitive computer work, the sensible next step is a measured trial, not a wholesale switch. Use GPT-6.1 Terra where speed and cost per task matter most, keep Sol and Astra for the hardest work, and judge GPT-6.1 Terra on finished work rather than launch-day numbers. That is the verdict the evidence supports today.
Sources and Attribution
- OpenAI announcement post, “GPT-6.1 Terra: delivering fast and capable computer use at a lower price,” October 26, 2026, via the OpenAI Developer Community. Launch wording, availability tiers, and benchmark claims as stated by OpenAI.
- BleepingComputer, “OpenAI launches GPT-6.1 Terra with 3x faster AI agents, 50% lower pricing,” October 2026. Pricing, Agents’ Last Exam result, and launch wording, cross-checked against SiliconANGLE and FoneArena coverage of the same announcement.
- SiliconANGLE, “OpenAI launches GPT-6.1 Terra for faster, better-performing AI agents at half the price,” October 26, 2026. Model positioning and Altman’s announcement post.
- Sam Altman, post on X, October 26, 2026, as quoted in multiple launch reports: “6.1 Terra is here! It is good at a lot of things, but we are especially excited for this new era of computer use.”
- FoneArena, “OpenAI introduces GPT-6.1 Terra with faster computer use and lower API pricing,” October 2026. Availability date, pricing qualifications, and prompt-injection resistance.
- Data Phoenix, “OpenAI launches GPT-6.1 Terra for computer use,” October 27, 2026. Terminal-Bench 4.0 figures, error-rate figures, pricing qualifications, and safety addendum summary.
- Artificial Analysis Intelligence Index results via Van Data Team, October 26, 2026. Independent scores of 48.1 against 42.1 for GPT-5.6 Terra, and per-task cost comparison.
- Amazon Web Services, GPT-6.1 Terra on Amazon Bedrock announcement, via Oossa, October 26, 2026. Bedrock availability and use-case examples.
External references: OpenAI announcement post and BleepingComputer launch report.
