Microsoft has launched seven of its own AI models — branded “MAI” — in a clear move to lean less on OpenAI and cut costs for developers building on Azure. The lineup was unveiled at the company’s Build 2026 developer conference.
The flagship is MAI-Thinking-1, Microsoft’s first reasoning model, alongside MAI-Code-1-Flash, which turns plain-English descriptions into working source code for apps and websites. Microsoft also showed MAI-Image-2.5, MAI-Voice-2 and MAI-Transcribe-1.5, rounding out a full multimodal stack.
Why it matters
The pitch is economics. By running these models on its own Azure infrastructure instead of licensing them, Microsoft avoids royalty payments and says it can pass the savings on. Microsoft AI chief Mustafa Suleiman claimed the new models beat OpenAI’s GPT‑5.5 on McKinsey’s benchmarks while costing roughly a tenth as much — and the company has reportedly capped revenue‑sharing with OpenAI and dropped its exclusive right to resell OpenAI’s models.
For anyone shipping software, MAI‑Code‑1 is the one to watch: a cheaper code model baked into the Microsoft stack changes the math on AI coding assistants. If you’re deciding which one to actually build with, our hands‑on guide to the best AI coding tools breaks down where each lands on speed, accuracy and price.
Source: CNBC.
AI Tools Worth is independent and unsponsored. Some of our linked guides contain affiliate links — they never change our verdicts.