Microsoft's first speech-to-speech model, MAI Realtime 1, is built for enterprise voice applications and priced at $0.01 per audio minute, the company said on November 171.
Microsoft says the model outperforms OpenAI's GPT Realtime 2.1 on quality, speed, and price across public benchmarks and Microsoft's own product evaluations. MAI Realtime 1 sits on the performance-price Pareto frontier for both the Big Bench Audio speech reasoning benchmark and the Tau-Voice agentic task completion benchmark, according to Microsoft.
Microsoft says it is bringing MAI Realtime 1 to Dynamics 365 Contact Center for customer service and to Dragon Copilot to support AI-simulated patient conversations that help nurses practice clinical interactions. Agoda and Reservations.ai are reportedly already testing the model.
MAI Realtime 1 is compatible with the OpenAI Realtime API. Reasoning is billed separately at $1 per million input tokens and $5 per million output tokens. Preview customers will not be billed until public preview begins, according to Microsoft.
The launch arrived the same day OpenAI released GPT-6.2 Astra and GPT-6.2 Sol, which paired benchmark gains with a 20% reduction in API pricing for its frontier model[1].
ANALYSIS Microsoft's decision to price MAI Realtime 1 at a cent per audio minute while claiming benchmark superiority over GPT Realtime 2.1 puts direct cost pressure on OpenAI's own real-time voice offering, extending the pricing competition between the two companies from text into speech.