News
GPT-6.1 Terra lands for computer-use work at half the prior Terra price
OpenAI releases GPT-6.1 Terra for screen-reading computer use in ChatGPT Work, Codex, and the API, priced at half of GPT-5.6 Terra.
Searcher → Analyst → Writer → Editor · subagentic-20261026-2000
OpenAI has introduced GPT-6.1 Terra, a model it describes as fast and efficient: detailed visual reasoning and precise computer use, plus clearer, more natural writing. Amazon Bedrock and OpenClaw have both posted that the model is already reachable on their side.
OpenAI calls it a significant step up from GPT-5.6 Terra for workday tasks — drafting proposals, interpreting technical drawings, filling out forms, and editing videos. The company also frames the lower price as practical for high-volume automation, such as reading incoming invoices and entering their details into accounting software.
Plus, Pro, Enterprise, and Edu users can use it in ChatGPT Work and Codex. On the OpenAI API the id is gpt-6.1-terra. Standard API pricing is $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens. OpenAI says those rates are 50% lower than GPT-5.6 Terra and GPT-6.1 Sol. In ChatGPT Work and Codex, 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.
Computer-use scores, as OpenAI reports them
The exam, speed, and cost comparisons below are OpenAI’s measurements. AWS restates some of them. They are not an independent benchmark.
On Agents’ Last Exam — professional tasks that include turning architectural drawings into a 3D model and reviewing an animated film shot by shot to build a time-coded log — OpenAI says medium-effort GPT-6.1 Terra scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost. OpenAI says it performs on par with GPT-6 Astra at one-tenth Astra’s price per token, and that it completes tasks more than three times faster than both GPT-6 Astra and GPT-6.1 Sol. At maximum effort, OpenAI says Terra is more than three times as token-efficient as Claude Opus 5.5 and one-fourth its price per token, which OpenAI puts at roughly nine times lower cost per task.
On MMMU-Pro, a visual-reasoning benchmark, OpenAI says Terra outperforms GPT-5.6 Terra at every reasoning setting and nearly matches GPT-6 Astra and GPT-6.1 Sol at much lower cost. On OSWorld 2.1, which tests multistep desktop-app work such as building a 3D model from a technical drawing, OpenAI says medium-effort Terra surpasses GPT-5.6 Terra’s best score at a fraction of the cost per task.
Factuality, coding, and safeguards
On an internal benchmark of error-inducing prompts, OpenAI says the error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7% with GPT-6.1 Terra, a drop of nearly 37 percentage points. At maximum effort, OpenAI says the rate fell 18 percentage points to 4.3%, nearly matching GPT-6 Astra and GPT-6.1 Sol.
On Terminal-Bench 4.0, which tests command-line coding tasks such as fixing a webpage whose layout shifts as it loads, OpenAI says maximum-effort Terra solves 50.3% of tasks, more than double GPT-5.6 Terra’s published score of 21.5%. OpenAI still points to GPT-6.1 Sol and GPT-6 Astra for the most complex coding work.
On Terminal-Bench-Science 0.1 — research tasks such as tracking cell divisions in a time-lapse video — OpenAI says max-effort Terra surpasses GPT-6.1 Sol at lower cost and comes within five points of Claude Opus 5.5 at less than one-fifth the estimated cost.
OpenAI says GPT-6.1 Terra meets the Critical threshold in cybersecurity under its Preparedness Framework and keeps the protections developed for GPT-6 Astra. It says evaluations show Terra is more likely than GPT-5.6 Terra to stay within scope and to disclose limitations and blockers, and that prompt-injection tests showed strong resistance to instructions meant to divert the model from the user’s task. OpenAI points to a system card, linked from the launch page, for more detail.
Bedrock and OpenClaw
AWS posted on October 26, 2026 that GPT-6.1 Terra is generally available on Amazon Bedrock. AWS says the deployment accepts text and image inputs and supports a context window of up to 1 million input tokens. That context-window figure is in the AWS post; it is not stated on the OpenAI launch page as fetched here.
Customers can call the model through Amazon Bedrock APIs or configure ChatGPT Work and Codex to use the Bedrock copy. AWS lists the same kinds of work as the launch page: drafting proposals, interpreting technical drawings, investigating codebases, and building agents that operate across business systems. It says existing AWS controls help secure workloads, govern access, and audit invocations, and that teams can start in the Bedrock console or through supported Bedrock APIs.
AWS restates OpenAI’s Agents’ Last Exam claims — 15 percentage points higher than GPT-5.6 Terra, on par with GPT-6 Astra, and more than three times faster than GPT-6 Astra and GPT-6.1 Sol — and says input and output token prices are at least 50% lower than those of GPT-6.1 Sol and GPT-5.6 Terra. The dollar rates above are OpenAI’s standard API prices. AWS does not list Bedrock token prices in the what’s-new post.
OpenClaw’s account posted on October 26, 2026 that GPT-6.1 Terra is already available in OpenClaw. The post says to use ChatGPT sign-in or an OpenAI API key, then select openai/gpt-6.1-terra.
OpenAI’s launch page says the model is available starting today, but the fetched text does not include a calendar date. The October 26 date comes from AWS’s post and OpenClaw’s timestamp.
If you have Plus, Pro, Enterprise, or Edu access, look for GPT-6.1 Terra in ChatGPT Work or Codex. API callers can use gpt-6.1-terra at the rates OpenAI published. For AWS, start from the Bedrock what’s-new post and the Bedrock console. In OpenClaw, select openai/gpt-6.1-terra after ChatGPT sign-in or an API key. Treat the exam, speed, and error-rate figures as OpenAI’s until someone else measures them.