OpenAI releases GPT-6.1 Terra for fast computer use

OpenAI released GPT-6.1 Terra on 26 October 2026 and made it available in the API as gpt-6.1-terra. The API changelog describes it as a fast, efficient model for visual reasoning, precise computer use and clear, natural writing. It can be called through the Responses and Chat Completions endpoints.

Specifications and pricing

According to OpenAI’s model page, GPT-6.1 Terra accepts text and image input and produces text output. Audio and video are not supported. The listed limits are:

• a 1,050,000-token context window and up to 128,000 output tokens;
• a knowledge cutoff of 1 June 2026;
• reasoning effort settings of none, low, medium (the default), high, xhigh and max.

Standard pricing is $1 per million input tokens and $5 per million output tokens, with cached input at $0.05 per million. Prompts above 272K input tokens are billed at twice the input rate and 1.5 times the output rate for the whole request. Batch and Flex pricing are 50% lower than Standard, and Fast mode costs twice the Standard price.

What developers should know

Tool calling requires the Responses API. The model page lists web search, file search, code interpreter, computer use, MCP and several other tools as supported there. Realtime, fine-tuning and embeddings endpoints are not supported.

OpenAI’s GPT-6 guide suggests starting at medium reasoning effort and raising it only if the quality gain justifies the extra time and tokens. For teams moving from GPT-5.6 Terra, it advises comparing both models on representative images and workflows at the same reasoning effort. It also says tool-using Chat Completions requests should move to the Responses API.

Why it matters

This is our analysis. Terra sits in the “fast and efficient” tier of OpenAI’s GPT-6 line, as OpenAI labels it, and its stated focus on screens and images makes it relevant to teams building agents that operate software. OpenAI has not published a comparison with other models on the pages we read, so claims about how it performs against GPT-5.6 Terra or competitors should be tested on your own tasks.