OpenAI released GPT-6.1 Terra on October 26, targeting visual reasoning, computer use and everyday writing. Its practical focus is work that moves from understanding an image to acting inside an application. The launch is confirmed in OpenAI’s API changelog.
Key developer details
- Visual input: interpret technical drawings, match product photos and work from application screens.
- Large context: a 1,050,000-token context window, with up to 128,000 output tokens.
- Standard pricing: $1 per million input tokens, $0.05 for cached input and $5 for output. Requests above 272,000 input tokens carry higher rates; tools and service options can add costs.
- API access: available as
gpt-6.1-terra; tool calling requires the Responses API.
The model’s context capacity and its lowest listed token rates are different limits. A workflow that fits inside the context window can still cross the long-context pricing threshold.
Why it matters for builders
Terra gives builders another option for bounded workflows: interpreting screenshots, updating records and making scoped code fixes. OpenAI recommends comparing it with Sol on identical tasks and checking the saved result—not just the completion message.
For SunMarc App Labs, a useful first test would be screenshot-based app and website layout checks, measuring missed defects, time and cost before adoption. A fixed set of screens and known issues would make the comparison repeatable; human review would determine whether the findings are useful.
This is a proposed evaluation, not an announced SunMarc feature or a claim of tested performance. As with our coverage of Kimi K3.1 Pro for longer coding tasks, the useful measure is the accepted result of the whole workflow, including review and retries.