Skip to content

Microsoft ships MAI Realtime 1, a speech model at $0.01 a minute

Microsoft's first speech-to-speech model skips the text step entirely and says it beats GPT-Realtime 2.1 on its own benchmarks, without publishing the scores.

Source: Microsoft

Microsoft announced MAI Realtime 1 at Ignite 2026 on 17 November, its first speech-to-speech model. It takes audio in and answers with audio directly, with no text transcription step in between.

Microsoft is pricing it at $0.01 per audio minute, and says it outperformed OpenAI’s GPT-Realtime 2.1 across external and Dynamics 365 benchmarks, though it has not published the actual scores behind that claim. The model is live to try now in MAI Labs, with wider access by registering interest, Microsoft’s usual path for a model still in early rollout.

Why it matters: a typical voice agent chains speech-to-text, a text model, and text-to-speech, paying latency and cost at every handoff. A native speech-to-speech model at a cent a minute gives anyone building a voice agent, phone bot or dictation tool a cheap new baseline to benchmark against whatever they run today.

The caveat: the “outperformed” claim comes from Microsoft’s own evaluation, with no published numbers and no independent benchmark to check it against yet.

Share

Related reading

More from The Wire

All briefs