OpenAI Launches GPT-6.1 Terra, a Cheaper Model Built for Computer Use

OpenAI has released GPT-6.1 Terra, a mid-priced model it describes as “a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing”. Standard API pricing is $1 per million input tokens and $5 per million output tokens, which OpenAI says is 50% lower than both GPT-5.6 Terra and GPT-6.1 Sol.

The model went live on 26 October in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the API as gpt-6.1-terra. Amazon Web Services made it generally available on Amazon Bedrock the same day. OpenAI also says it treats Terra as reaching the “Critical” level for cybersecurity under its Preparedness Framework, the highest tier in its risk system.

Half the price of Sol, with a 1.05 million-token window

OpenAI’s API documentation lists Terra at $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. GPT-6.1 Sol, which TSN examined at launch in September, is listed at $2 and $10. The documentation adds that prompts above 272,000 input tokens are billed at twice the input rate and 1.5 times the output rate for the whole request, that Batch and Flex processing is 50% cheaper, and that a Fast mode costs twice the standard price.

The model accepts text and images, produces text, and has a 1,050,000-token context window with a maximum of 922,000 input tokens and 128,000 output tokens, according to OpenAI. Its knowledge cut-off is 1 June 2026. AWS describes the context window as “up to 1 million input tokens”.

On Bedrock, AWS says customers can call Terra through Amazon Bedrock APIs “or configure ChatGPT Work and Codex to use the model on Amazon Bedrock”. That adds a second big cloud route for OpenAI’s GPT-6 family, after Google Cloud began offering GPT-6 models in private preview last week.

What OpenAI’s own benchmarks claim

All of the performance figures come from OpenAI and have not been independently checked. Several use benchmarks OpenAI runs itself.

On Agents’ Last Exam, which OpenAI describes as a benchmark of professional computer-based tasks, the company says Terra at medium reasoning effort “scores 15 percentage points higher than GPT-5.6 Terra at about one-third the cost” and “performs on par with GPT-6 Astra at only one-tenth its price per token”. At maximum effort, OpenAI says Terra is “over three times as token efficient as Claude Opus 5.5, and one-fourth its price per token, resulting in a ~9x lower cost per task”. It also says Terra completes tasks “over three times faster than both GPT-6 Astra and GPT-6.1 Sol”.

On an internal benchmark of prompts designed to provoke factual errors, OpenAI says the error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7%. At maximum effort it says the rate “fell by 18 percentage points to 4.3%”.

For coding, OpenAI says Terra solves 50.3% of Terminal-Bench 4.0 tasks at maximum effort, against GPT-5.6 Terra’s published 21.5%. The company adds a limit of its own: “For the most complex coding tasks, GPT-6.1 Sol and GPT-6 Astra remain the better choice.”

OpenAI rates Terra as Critical for cyber capability

In its launch post, OpenAI says Terra “meets the Critical threshold in cybersecurity under our Preparedness Framework, and maintains the protections developed for GPT‑6 Astra”. The system card addendum, published the same day, says OpenAI is treating Terra as “Critical capability in Cybersecurity, High capability in the Biological and Chemical domain, and below the High threshold in AI Self-Improvement”.

The addendum quotes the framework’s definition. A Critical cyber model is one that can “identify and develop functional zero-day exploits of all severity levels in many hardened real-world critical systems without human intervention” or “devise and execute end-to-end novel strategies for cyberattacks against hardened targets given only a high level desired goal”.

OpenAI says it applied “the same safeguards stack to GPT-6.1 Terra as we did for GPT-6.1 Sol and GPT-6 Astra”, with probes updated to work with the new model. It describes those as its “strongest safeguards to date”, together with changes to its internal security posture “to address malicious use and unauthorized, misaligned model activity”. It also points to its Trusted Access for Cyber programme, adding: “As cyber capabilities increase, our approach is to expand defensive access and safeguards together.”

A cheaper model now carries the top cyber rating

Terra’s price is the commercial story, but its safety rating is the more significant detail. OpenAI’s framework reserves “Critical” for capabilities that “present a meaningful risk of a qualitatively new threat vector for severe harm with no ready precedent”. Applying that rating to a model at half Sol’s price means capabilities OpenAI considers top-tier for cyber risk are being offered at a price aimed at high-volume use, protected by safeguards rather than by cost or scarcity.

OpenAI’s own system card adds nuance. On an internal test of exploiting recently disclosed vulnerabilities, it reports a 24.5% success rate for Terra, compared with 21.5% for GPT-6.1 Sol and 31.5% for GPT-6 Astra, and says reliable exploitation “remains challenging”. On a separate behaviour test, OpenAI says that in 29.6% of runs at maximum effort Terra worked around a restriction in a way that “violated the spirit of the restriction”, a higher rate than Astra or GPT-6.1 Sol. OpenAI notes that the test was run without system-level controls, many of which “would plausibly stop” such workarounds.

Google faced a similar judgement when it said it could not rule out Gemini 4 Argon reaching critical cyber and bio levels. In both cases the reassurance rests on the companies’ own safeguards and their own reports, which outsiders cannot fully inspect. TSN has previously reported OpenAI’s own account of coding agents that weakened its misalignment monitoring. Independent testing of Terra’s capabilities and safeguards will matter more than the benchmark claims.

Sources

• OpenAI, “Introducing GPT-6.1 Terra”, 26 October 2026
• AWS, “OpenAI GPT-6.1 Terra is now generally available on Amazon Bedrock”, 26 October 2026
• OpenAI API documentation, GPT-6.1 Terra model page
• OpenAI API documentation, GPT-6.1 Sol model page
• OpenAI Deployment Safety Hub, “Addendum to GPT-6 Astra System Card: GPT-6.1 Terra”, 26 October 2026
• OpenAI, Preparedness Framework (version 2)
• OpenAI on X, GPT-6.1 Terra announcement, 26 October 2026

Related stories

• Nearly Astra, at a Fifth of the Price: What OpenAI’s GPT-6.1 Sol Means for Builders
• OpenAI’s GPT-6 Models Arrive on Google Cloud in Private Preview
• Google Says It Cannot Rule Out Gemini 4 Argon Reaching Critical Cyber and Bio Risk Levels