Microsoft introduced seven new in-house “MAI” AI models, a clear move to reduce its dependence on OpenAI and cut costs for developers building on Azure.
The models that matter
- MAI-Code-1-Flash: turns plain-language descriptions into working source code for apps and websites.
- MAI-Thinking-1: a reasoning model tuned for efficiency.
- Five more: rounding out a family aimed at different cost/performance points.
Why Microsoft is doing this
Owning the models lowers per-token costs, reduces reliance on a single partner, and lets Microsoft tune efficiency for Copilot and Azure workloads.
What this means for you
- Copilot users may see faster, cheaper responses as MAI models take over routine tasks.
- Developers on Azure get more model options at different price points.
- The frontier (GPT-5.5, Claude, Gemini) still leads on the hardest tasks — MAI targets efficiency, not the top of the benchmark charts.
FAQ
Does this replace OpenAI in Microsoft products? Not fully — it’s a hybrid strategy, using MAI where it’s cost-effective.