What was announced
OpenAI released GPT-6.1 Terra on October 26, positioning it for writing, interpreting images, and working across software. The developer changelog lists the model as gpt-6.1-terra for the Responses and Chat Completions APIs. [1][4]
OpenAI reports improved visual reasoning and computer-use performance compared with GPT-5.6 Terra, alongside lower standard API prices. Those are provider evaluations, rather than Tinker & Tokens tests. Benchmark gains do not establish that it will finish your particular job correctly or more cheaply. [1]
Check the product and the tools
ChatGPT Learn lists Terra in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise, and Edu plans. It is not available in ordinary Chat. Look for the exact model in Advanced; enterprise availability depends on workspace and administrator controls. Browser and computer use have their own setup and access requirements. [2]
For developers, tool calling requires the Responses API. Native model support is text input/output and image input; audio and video are not supported. A model editing a video through software tools is different from native video input. The model name alone does not supply an application with working tools or permission to act. [3]
Read the price with its conditions
Standard API text prices per one million tokens are $1 for input, $0.05 for cached input, and $5 for output. Cache writes and tool use can add charges; processing tiers and regional options change pricing. For prompts above 272,000 input tokens, 2× input/cache rates and 1.5× output rates apply to the full request. These API rates are separate from Work or Codex plan allowances. [1][3]
Our take
For a fictional neighborhood shop, start with three supplier product photos and a short catalog. Ask Terra to draft a list of mismatched labels, with no changes to the live store. Supply copies of only the relevant files, define a successful match, and check the selected app, model, and permissions before starting.
Review every suggested match against the originals before approving edits. Record time spent checking and correcting alongside model and tool charges. A lower token price can help, but retries, mistakes, and human review affect the cost of a finished task. Use that small comparison to decide whether Terra earns a place in your workflow.