A cheaper model becomes useful when it can finish the task on the screen. OpenAI’s new Terra release makes that a more practical comparison for teams automating ordinary application work.
In its October 26 announcement, OpenAI positions GPT-6.1 Terra around visual understanding and computer use. Standard API rates are $1 per million input tokens, $0.05 for cached input and $5 for output. The company says those rates are 50% below GPT-5.6 Terra and GPT-6.1 Sol. It lists availability in the API as gpt-6.1-terra and in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users.
OpenAI reports gains on computer-use and coding evaluations. On Agents’ Last Exam, it says Terra at medium reasoning effort scores 15 percentage points above GPT-5.6 Terra at about one-third the cost. These are the vendor’s measurements under its evaluation conditions, rather than evidence of a saving on your particular process.
The accompanying system card makes an important qualification: evaluations used research or API environments, whose prompts, tools and reasoning settings can differ from production ChatGPT. It also says some comparison values for earlier models reflect later versions. A model name alone therefore does not identify everything being compared.
Consider a hypothetical invoice-entry trial. Give each candidate the same documents, application permissions and completion rule. Count a result only when the correct fields are entered and the required confirmation is preserved. Keep the human corrections, retries and elapsed time beside the token bill. This proposed comparison would test the complete workflow, including the points where screen interpretation becomes an action.
A lower rate gives a team room to run that experiment. The deployment decision still rests on whether the agent completes the permitted work and leaves enough evidence for someone to check it.