Microsoft announces MAI-Thinking-1 and MAI-Code-1-Flash, two new text LLMs with 35B and 5B parameters respectively

MAI-Thinking-1 (35B parameters) is available to "select early partners" and is described as preferred to Sonnet 4.6 in Microsoft's blind human side-by-side evaluations. MAI-Code-1-Flash (5B parameters) is rolling out to GitHub Copilot individual users in Visual Studio Code and is "purpose-built for GitHub Copilot and VS Code to deliver high performance and lower cost."

Both models were trained by Microsoft from the ground up using "enterprise grade, clean and commercially licensed data" without distillation from third-party models, according to the company.

The low parameter counts (35B and 5B) stand out as notably smaller than many currently available large models, though the article notes the author has not yet tried either model.