GPT-6.1 Terra introduces a lower-priced computer-use model

What changed

Lower token prices do not guarantee lower total workflow cost: completion rates, retries, tool use and oversight still matter.

The price comparison concerns standard API token rates. It does not measure the total cost of completing a business workflow. A lower rate per token can be offset by a longer interaction, more retries, additional tool calls or a higher need for human review. Compare a complete task rather than assuming the published price ratio will also be the cost ratio for every use case.

Computer-use and cross-app tasks involve both reasoning and interaction with software. A useful evaluation therefore checks whether the model reaches the intended result, handles intermediate states correctly and preserves the required boundaries around account actions. A fluent explanation of a task is not the same as a correctly completed task.

For a marketing workflow, start with a repeatable, bounded example such as reconciling a report or preparing a reviewable document. Define the expected output and inspect errors as well as successes. Keep consequential actions, such as campaign changes or publication, under the appropriate human authorisation process rather than treating model availability as permission to automate them.

The announcement positions Terra for these tasks; it does not establish that every existing integration supports the model or that a particular advertiser's workflow will perform better after switching. Check product access, integration requirements and the measured outcome of the test separately.

OpenAI's 26 October announcement is the primary source, with James Massey retained for additional contribution in the newsletter. This is a model and pricing update, not a vendor recommendation or a guarantee of lower overall operating cost.