OpenAI introduced GPT-6.1 Terra on October 26, 2026, adding a model aimed at visual reasoning and computer use. The API changelog records the release, and the launch announcement lists standard pricing of $1 per million input tokens, $0.05 per million cached input tokens, and $5 per million output tokens.
Developers can select gpt-6.1-terra through the API. OpenAI also says the model is available to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex. Subscription access and metered API spending are different purchasing routes.
What the prices mean for a small workload
Consider an illustrative workload with one million input tokens and 100,000 output tokens. At the published standard rates, ordinary input costs $1, and output costs $0.50, for a $1.50 token bill. If all of that input qualifies for the cached rate, the same token counts cost $0.55.
| Illustrative input treatment | Input cost (USD) | Output cost (USD) | Token total (USD) |
|---|---|---|---|
| 1 million ordinary input tokens | $1.00 | $0.50 | $1.50 |
| 1 million eligible cached input tokens | $0.05 | $0.50 | $0.55 |
The cached input rate is 95% below the ordinary input rate. That percentage applies only to the input component, so calling it a 95% reduction in the whole workflow bill would be wrong. In this example, output alone still costs ten times as much as the cached input.
Lower token prices need a task-level comparison
OpenAI describes the standard API prices as 50% below GPT-5.6 Terra and GPT-6.1 Sol. Its reported computer-use and coding improvements are company evaluations; MustHave has not independently reproduced them. The accompanying deployment safety material provides further context on the release.
For a browser or desktop agent, a useful comparison would hold the task and success criteria constant, then record tokens, tool calls, retries, and human corrections. A cheaper attempt that needs several retries can cost more per completed task. Conversely, a higher output bill may still be worthwhile if it produces a usable result sooner.
The new rates extend the pricing changes covered in our earlier GPT-5.6 pricing report. The arithmetic above is a reproducible budget example, not a measured agent run or a quote for a complete production workload.
