OpenAI introduces GPT-6.1 Terra, a faster and cheaper computer-use model

What OpenAI released

OpenAI introduced GPT-6.1 Terra on 26 October 2026 as a fast, efficient model that combines detailed visual reasoning and precise computer use with clearer, more natural writing. The announcement describes it as a significant step up from GPT-5.6 Terra, aimed at the high-volume work that fills a workday: drafting proposals, interpreting technical drawings, filling out forms and editing videos.

Terra is the latest in the GPT-6 family, sitting below the flagship GPT-6 Astra and GPT-6.1 Sol as the cost-focused option. OpenAI frames it as a cost-intelligence play: for developers automating high-volume tasks like reading invoices and entering them into accounting software, the lower price changes the math.

The pricing

Standard API pricing is $1 per million input tokens and $5 per million output tokens, with cached input at $0.05 per million tokens. OpenAI says that is 50% lower than GPT-5.6 Terra and GPT-6.1 Sol.

In ChatGPT Work and Codex the same usage allowance covers more tasks with GPT-6.1 Terra than with GPT-5.6 Terra, and substantially more than with GPT-6.1 Sol or GPT-6 Astra, according to the company. BleepingComputer reported the model is also generally available on Amazon Bedrock the same day, which gives AWS customers a managed route to it without writing directly to OpenAI's API.

The price point matters for the automation use case OpenAI is selling. Reading invoices, entering details into accounting software and similar high-volume jobs were previously the turf of cheaper, smaller models. Terra is OpenAI's attempt to bring a GPT-6-series model down into that range while keeping the visual grounding that document work needs.

The benchmark claims

OpenAI's own numbers are the source for most of the comparisons. On Agents' Last Exam, a benchmark of professional tasks that includes turning architectural drawings into a 3D model and reviewing animated film, GPT-6.1 Terra at medium reasoning effort scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost, and OpenAI says it performs on par with GPT-6 Astra at only one-tenth its price per token.

The company also says Terra is much quicker, completing tasks over three times faster than GPT-6 Astra and GPT-6.1 Sol, and three times more token efficient than Claude Opus 5.5 at max effort, at one-fourth the price per token. On the error-rate benchmark OpenAI uses for factuality, the low-effort error rate dropped from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra, and at max effort to 4.3%. These are all OpenAI's internal measurements, not independent results.

On Terminal-Bench 4.0, which tests command-line coding, GPT-6.1 Terra solves 50.3% of tasks at maximum reasoning effort, more than double GPT-5.6 Terra's published 21.5% score. On OSWorld 2.1, a benchmark for desktop agent tasks, Terra surpasses GPT-5.6 Terra's best score at a fraction of the cost per task.

None of these results have been independently reproduced, and OpenAI's benchmarks are not standardized against each other the way public leaderboards would be. The useful comparison is between Terra and the older models OpenAI itself tested, where the direction of the change is clear even if the exact margins should be treated as vendor-reported.

How to use it

GPT-6.1 Terra is available to all Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex today. Developers can call it through the OpenAI API as gpt-6.1-terra.

AWS announced the model was generally available on Amazon Bedrock the same day, so enterprises running on AWS can invoke it there. OpenAI positions the model for agentic work in the browser and desktop arena, directly competing with the computer-use models that have become a category focus this year.

Safety and limits

OpenAI says GPT-6.1 Terra meets the Critical threshold in cybersecurity under its Preparedness Framework and maintains the protections built for GPT-6 Astra. The system card addendum is published on OpenAI's Deployment Safety Hub.

The practical limit is the same one that applies to all computer-use models: benchmarks measure single tasks, and real desktop automation is messier. OpenAI recommends GPT-6.1 Sol and GPT-6 Astra for the most complex coding jobs, so Terra is explicitly not the top reasoning model. Its value is speed and token cost for repetitive work.

What to watch

OpenAI is using Terra to defend the high-volume agent market against open-weight models that have competed on price. The $1/$5 per million token price puts a GPT-6-series model in the same range as the cheapest frontier-tier offerings.

Watch for third-party benchmarks of computer-use performance, and whether GPT-6.1 Terra holds the claimed advantage on OSWorld 2.1 when independently tested. The model also lands in a crowded field of computer-use releases, so real-world adoption will decide whether the price drop matters.