OpenAI documents GPT-6.1 Terra with image input and tool use

OpenAI describes GPT-6.1 Terra as a fast, efficient model for visual reasoning, precise computer use and natural writing.[1] Its examples include matching product photos to catalog entries, interpreting technical drawings, acting on application screens, and drafting or revising text.[1] The listed inputs are text and images, while the listed output is text.[1] OpenAI says the model does not support audio or video, a relevant boundary for teams evaluating it for workflows that depend on those media.[1]

OpenAI lists a context window of 1,050,000 tokens, a maximum input length of 922,000 tokens and a maximum output length of 128,000 tokens.[1] Those are separate limits, so teams should check both the amount of material they send and the output they expect when planning a workload. OpenAI gives June 1, 2026, as the model’s knowledge cutoff.[1] Reasoning effort can be set to none, low, medium, high, xhigh or max, with medium as the default.[1]

OpenAI lists the Chat Completions, Responses and Batch endpoints as supported.[1] Tool calling with GPT-6.1 Terra requires the Responses API, so teams considering application actions should account for that requirement.[1] The listed standard rates are $1 per million input tokens and $5 per million output tokens; cached input costs $0.05 per million tokens.[1] OpenAI says Batch and Flex are priced 50% below Standard, Fast mode costs twice the Standard rate, and regional processing carries a 10% premium where available.[1]

Lumina reported that GPT-6.1 Terra appeared in 18 evaluation configurations, with 17 configurations changing from GPT-5.6 Terra to GPT-6.1 Terra.[2] Lumina also reported that the pull request was merged on October 16.[2] In its description of the setup, Lumina said Sol handles the task while Terra is configured to judge the work.[2]