OpenAI introduces GPT-6.1 Terra

The release makes the model available to Plus, Pro, Enterprise and Edu users of those applications. Standard API charges are $1 per million input tokens, $0.05 for cached input and $5 for output. OpenAI illustrates the intended work with reading invoices and entering their details into accounting software. These are provider examples. The company still recommends Sol or Astra for the most complex coding tasks,.

The accompanying evaluation document explains an important boundary around factuality claims. Its error-focused test uses anonymized conversations that users had flagged for mistakes. That deliberately difficult selection does not measure the error rate people should expect in ordinary production use. The test was refreshed, and earlier models were rerun; its results are not directly comparable with older system cards. It checks both whether a response contains any error and whether a particular reported mistake recurs. Research or API evaluation conditions can also differ from ChatGPT in prompts, tools and reasoning effort.

The Decoder’s report by Matthias Bastian, published October 26, places the launch in the market for less expensive models used for practical work. It combines the vendor announcement with Artificial Analysis evaluations. That account is secondary coverage; the product details above come from the provider, and the evaluation caveats from its technical documentation.