GPT-6.1 Terra turns up in Microsoft's eval configs as The Information flags new OpenAI, Anthropic and Google models

An id for GPT-6.1 Terra, a model OpenAI has not announced, is already in Microsoft's code. A pull request in the dotnet/efcore repository, the .NET team's Entity Framework Core project, merged on October 16, switched 18 evaluation configs from gpt-5.6-sol with a gpt-5.6-terra judge to gpt-6.1-sol with gpt-6.1-terra as the judge model. The configs run through GitHub's Copilot SDK. @LuminaBench spotted it on X on October 19; we read the pull request (#39217) and the config files directly.

WHAT IT SHOWS, AND WHAT IT DOES NOT

An id in a config file is not a launch. It shows that a Microsoft team set its evals up to call a model named gpt-6.1-terra three days ago, through the Copilot SDK. It does not show a price, a date, or that anyone outside Microsoft and GitHub can use it.

THE INFORMATION'S SCOOP

The same day, The Information's Stephanie Palazzolo reported that Anthropic is preparing to release a new model that is better in reasoning and intuition than its current ones, under the headline "Anthropic Readies Model That Reads Users' Intentions Better." In a post on X she said the newsletter also has news on Google's Gemini 4 Carbon and a potential new Terra model from OpenAI. The article is paywalled; we read its headline and summary, not the full text, and what we read names no Anthropic model and no date.

WHAT WE ARE WATCHING

Our hunts cover all three: a new GPT Terra, Fable 5.5, and Gemini 4's Barium and Carbon checkpoints. We widened the Terra hunt today, so a gpt-6.1-terra listing files under Terra rather than under the shelved GPT-6.1 flagship.

WHAT WE DO NOT KNOW

When any of them ships, what GPT-6.1 Terra will cost, and which Anthropic model The Information is describing.