Microsoft's new MAI models
Microsoft has introduced two new text LLMs, MAI-Thinking-1 and MAI-Code-1-Flash, aimed at enhancing performance and reducing costs. The MAI-Thinking-1 model features 35 billion parameters and is currently available to select early partners, while MAI-Code-1-Flash is designed for GitHub Copilot users. Both models are built using clean and commercially licensed data, raising questions about their training sources.
- ▪Microsoft announced two new text LLMs: MAI-Thinking-1 and MAI-Code-1-Flash.
- ▪MAI-Thinking-1 has 35 billion parameters and is available to select early partners.
- ▪MAI-Code-1-Flash is purpose-built for GitHub Copilot and VS Code, rolling out to individual users.
Opening excerpt (first ~120 words) tap to expand
Microsoft announced two new text LLMs this morning - MAI-Thinking-1 (reasoning, 35B parameters, available to "select early partners") and MAI-Code-1-Flash (5B parameters, "purpose-built for GitHub Copilot and VS Code to deliver high performance and lower cost [...] rolling out to GitHub Copilot individual users in Visual Studio Code"). I've not been able to try either of them just yet. It's very interesting to see Microsoft releasing models with such low parameter counts, especially given how expensive larger models are to access right now. They claim MAI-Thinking-1 "is preferred to Sonnet 4.6 in our blind human side-by-side evaluations", which is impressive for a 35B model seeing as I frequently run models larger than that on my own laptop.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Simon Willison's Weblog.
Discussion
0 commentsMore from Simon Willison's Weblog
