MAI Realtime 1: Microsoft's first live speech-to-speech model
Microsoft's first native speech-to-speech model — it listens and talks back without a separate transcript step — claiming to beat GPT-Realtime 2.1 with 98.9% on Big Bench Audio, 880ms median time-to-first-audio, from $0.01 per minute.
Why it matters
Microsoft's first native voice-conversation model, priced aggressively for voice-agent builders.
What you could build with it
A small team building a phone-based booking agent could use this model to hold natural spoken conversations with callers — confirming appointments and answering questions in real time — without stitching together separate transcription, language and speech-synthesis services.
Does it hold up?
Too early to judge: announced at Ignite on Nov 17; the accuracy and latency figures are Microsoft's own and no independent voice-agent deployments have surfaced yet.
Built with MAI Realtime 1
- Microsoft takes on OpenAI and Anthropic with seven new AI models at Ignite 2026neowin.net · Ignite launch coverage including MAI Realtime 1, with the vendor's accuracy, latency and pricing claims.
- Microsoft's Seven MAI Models at Ignite: Frontier Quality at Single-Digit Pricesdevopsby · Ignite recap: Realtime 1 is Microsoft's first speech-to-speech model at $0.01 per audio minute, with the honest caveat that all figures are vendor-reported.
- Microsoft's Seven New MAI Models: Cheaper AI Takes on Rivalstechorcoffee · Model-by-model breakdown; Realtime 1 is tryable in MAI Labs while Microsoft gathers feedback.
Learn more
First spotted on article: source.
More AI models
Edge0-145B: open 145B MoE that runs on a single consumer GPUEdge0 open-sourced 145B and 120B latent-MoE models derived from Qwen3.8-Flash-Next, plus an inference framework that…models · JEV 0.74Strata: run a 125B model on a gaming PCFree, MIT-licensed local inference engine that runs Qwen3.8-Flash-Next (125B MoE, about 6B active per token) on…models · JEV 0.73Qwen4-Max: Alibaba's new flagship model goes liveAlibaba's 2.4T-parameter flagship is now generally available via API, with a 1M-token context, native vision, and…models · JEV 0.67smolmaxxing: run 16B LLMs on a 1 GB TV box or a 2013 iPadAn MIT-licensed llama.cpp patch set that runs language models up to a 16B MoE on devices with about 1 GB of usable RAM…models · JEV 0.65Blue TTS v3: a 388MB open text-to-speech model with voice cloningAn open-source 388MB text-to-speech model (MIT) with voice cloning from about 10 seconds of speech, 7 languages, 15…models · JEV 0.65Haru-4BOpen full-duplex spoken-dialogue foundation model that listens and speaks simultaneously (7.5 Hz frames, 133 ms per…models · JEV 0.63
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.