Microsoft unveils MAI-Thinking-1 and MAI-Code-1-Flash, two new text LLMs. MAI-Thinking-1 boasts 1 trillion parameters with 35 billion active, claiming superior performance to Claude 3 Sonnet in blind evaluations, and is trained on clean, enterprise-grade data. MAI-Code-1-Flash, a 137 billion parameter model with 5 billion active, is purpose-built for GitHub Copilot to enhance performance and reduce costs.
Opening Kapyn…