Summary
Tip
See the Text and images tab for more details!
Microsoft AI develops capable, safe models for the demands of real work. In this module, you explored how the MAI portfolio addresses different input and output modalities.
You learned that:
- MAI-Thinking-1 supports chat and agent experiences that require multi-step reasoning, mathematics, coding, and enterprise analysis.
- MAI-Code-1.1-Flash supports efficient agentic coding workflows, including planning, implementation, debugging, and prototyping from visual designs.
- MAI-Voice-2 prioritizes natural, consistent speech, while MAI-Voice-2-Flash prioritizes low latency for interactive experiences.
- MAI-Transcribe-1.5 creates domain-aware transcripts across languages and challenging audio conditions.
- MAI-Image-2.5-Pro, MAI-Image-2.5, and MAI-Image-2.5-Flash provide different balances of visual capability, quality, and cost for image generation and editing.
The right model is the one that meets the quality, latency, cost, language, and safety requirements of your workload. Evaluate candidate models with representative inputs, measure the complete application experience, and review current model cards and service documentation before deploying to production.