OpenAI releases GPT-6.1 Terra with lower prices for computer-use AI

OpenAI has released GPT-6.1 Terra for computer-use tasks, with standard API prices half those of GPT-5.6 Terra. The model is available through the API and to Plus, Pro, Enterprise and Edu users via ChatGPT Work and Codex.

OpenAI Watch analysis

What happened

Terra is designed to interpret screen content, navigate applications and act across software workflows. OpenAI describes uses including filling in forms, drafting documents, interpreting technical drawings and editing video. EdTech Innovation Hub’s report says the standard API rates are $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens.

Our top picks

• Cheaper standard API use
  OpenAI lists standard prices 50% below GPT-5.6 Terra and GPT-6.1 Sol.
• More computer-use capability
  OpenAI says Terra scored 15 percentage points above GPT-5.6 Terra on Agents’ Last Exam at medium reasoning effort, at about one-third of the cost.
• Faster benchmark runs
  OpenAI reports that Terra completed benchmark tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol.
• Fewer factual errors in one internal test
  OpenAI reports a fall from 53.6% to 16.7% at low reasoning effort on its mistake-inducing prompts benchmark.
• A substantial coding-score increase
  On Terminal-Bench 4.0, OpenAI reports 50.3% for Terra against 21.5% for GPT-5.6 Terra at maximum reasoning effort.

Why it matters

The lower prices and stated improvements make this a potentially more practical option for repetitive work that involves operating software, not just answering questions about it. Developers can access it through the API, while eligible subscribers can use it in ChatGPT Work and Codex.

The performance figures above are OpenAI’s published evaluations, as reported by EdTech Innovation Hub. They are useful clues about the company’s intended workloads, not a guarantee of results in every workplace application.

Our read

Terra’s combination of lower prices and reported gains gives developers a concrete reason to test it on their own computer-use tasks. The best next step is not to trust a benchmark or a launch claim wholesale, but to compare cost, speed and error rates on the workflows that matter to you.

What to watch

• How Terra performs on real workplace tasks outside OpenAI’s evaluations.
• Whether its lower prices translate into lower total costs for longer workflows.
• How it compares with GPT-6 Astra and GPT-6.1 Sol on more complex coding work.

Discussion spark: For computer-use agents, should lower cost be enough to make a model worth adopting, or should developers wait for independent results on real workplace tasks?

OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.