OpenAI Launches GPT-6.1 Terra: Faster Computer-Use AI at Half the Price of Its Own Comparable Models

OpenAI has announced GPT-6.1 Terra, a model built for high-volume professional work that can read screens, control application interfaces and produce natural written output — all at a price point the company says undercuts comparable models in its own lineup by half.

The announcement came via OpenAI’s official channels, with the model available through the API under the identifier `gpt-6.1-terra`. It is also rolling out inside ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu subscribers, though availability depends on account type, region and workspace settings.

What GPT-6.1 Terra Actually Does

The model is designed around what OpenAI calls computer use — the ability to interpret what’s on a screen and take actions within software, not just answer questions about it. That means it can fill in forms, process invoices, draft proposals, read technical drawings and support video editing workflows without a human clicking through each step.

It’s part of OpenAI’s broader GPT-6 series, positioned as an efficiency-focused option rather than the most capable model in the range. Think of it as the workhorse rather than the thoroughbred — built for repetitive, high-volume tasks where speed and cost matter more than raw reasoning power.

Developers can tune how hard the model thinks. Reasoning-effort settings run from none through to max, with medium as the default — a practical dial that lets teams balance cost against depth of output depending on the task.

The Numbers

API pricing is set at $1 per million input tokens and $5 per million output tokens — around 80p and £4 respectively at current exchange rates, though OpenAI publishes prices in US dollars and no UK-specific pricing page was independently verified. Cached inputs drop sharply, to $0.05 per million tokens.

OpenAI says those standard prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol.

An Amazon Web Services model card adds some performance context. It reports that GPT-6.1 Terra scored 15 percentage points higher than GPT-5.6 Terra on OpenAI’s Agents’ Last Exam evaluation — a benchmark designed to test autonomous task completion. The same card states that Terra completed those evaluation tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.

Those figures come from OpenAI and AWS documentation rather than independent testing, so they should be read as the company’s own account of the model’s performance.

Who It’s For

GPT-6.1 Terra is not part of the standard ChatGPT experience that most consumers use day-to-day. It’s aimed squarely at developers and organisations building automated workflows — teams that process large volumes of documents, run screen-based administration at scale, or want to automate repetitive back-office tasks without paying for more expensive models.

OpenAI also notes that API usage is billed separately from ChatGPT subscriptions, which matters for any organisation budgeting for deployment.

The rollout date in OpenAI’s product documentation is listed as 9 April 2026, with a further update recorded between 26 and 27 October 2026. The exact chronology hasn’t been independently clarified.

The Broader Picture

Lower token costs and the ability to control software directly could make automation more viable for tasks that were previously too expensive or too fiddly to hand off to a model. But computer-use AI also raises real questions about oversight. A system that can click buttons, submit forms and move data between applications needs careful guardrails — above all where errors could have financial, legal or operational consequences.

OpenAI has published a system-card addendum for GPT-6.1 Terra covering safety considerations, though the specifics of those assessments weren’t independently reviewed for this article.

GPT-6.1 Terra is not the most powerful model OpenAI makes, but one priced and designed to run at scale inside real business processes.

The workforce question is live too. Automation of screen-based admin could reduce repetitive work for some employees while increasing demand for people who can supervise, audit and correct AI systems. Neither outcome is guaranteed, and the balance will depend heavily on how organisations choose to deploy the technology.

What This Means for Kent Residents

Organisations across Kent — including Kent County Council, Medway Council, NHS Kent and Medway ICB, the University of Kent and Canterbury Christ Church University — could potentially use a model like GPT-6.1 Terra to handle document drafting, invoice processing and screen-based administration at lower cost than previous AI tools. Any public-sector or health body considering deployment would need to carry out formal data-protection assessments and confirm that contractual and security arrangements meet UK GDPR requirements before processing personal or sensitive information. For private businesses in the county, the lower API pricing makes AI-assisted automation more accessible than it has been, though independent verification of the model’s real-world performance remains limited. No specific deployment by a Kent organisation has been confirmed.