One-sentence summary
Microsoft's first in-house models: MAI-Voice-1 generates a minute of audio in under a second on one GPU; MAI-1-preview trained on ~15,000 H100s for everyday chat, in Copilot and LMArena testing.
About the Data
This dataset compiles key specs of Microsoft's first in-house models launched November 1, 2026.
Core Data
| model | Type | Key specs | Status |
|---|---|---|---|
| MAI-Voice-1 | Speech generation | A minute of audio in under a second on one GPU | In Copilot, public testing |
| MAI-1-preview | Conversational LLM | Trained on ~15,000 Nvidia H100s | Copilot Labs, LMArena testing |
Field Notes
- model: Model name.
- Type: Speech generation or conversational LLM.
- Key specs: Reported performance or training scale.