I'm LongbridgeAI, I can summarize articles.Microsoft launched three new AI speech models: MAI-Transcribe-2-Streaming, which transcribes in 320ms and supports 60 languages; MAI-Voice-2.1 for high-quality speech; and the cost-effective Flash version. These models enhance real-time responsiveness for applications like customer service and voice assistants. Meanwhile, Wall Street analysts maintain a Strong Buy consensus on Microsoft (MSFT) stock with an average price target of $575.91, indicating 11.8% upside potential.
Software giant Microsoft (MSFT) introduced three new AI speech models on Thursday, led by MAI-Transcribe-2-Streaming, which can begin showing words roughly 320 milliseconds after they are spoken. For reference, competing transcription models often take more than 500 milliseconds. Microsoft says the new model supports 60 languages and automatically detects which language a person is speaking. Introductory pricing is $0.54 per hour through the end of 2026.
Interestingly, the model does not wait for someone to finish a sentence before producing text. Instead, it creates an early version of the transcript as speech arrives. It then improves that text as more context becomes available and locks in the final version once the person finishes speaking.
That could make voice-based AI feel more responsive. For example, a customer-service system could start figuring out why someone called before the sentence is complete. A voice assistant could also begin preparing its next action sooner.
As a result, Artificial Analysis currently ranks MAI-Transcribe-2-Streaming first for accuracy on both early and final transcripts, according to Microsoft. The new model builds on Microsoft's earlier MAI-Transcribe-2, which already supports 60 languages and is designed for real-world audio such as meetings and call-center work.
Microsoft Is Also Expanding Its AI Voice Models
Microsoft also introduced MAI-Voice-2.1 and MAI-Voice-2.1-Flash. MAI-Voice-2.1 costs $22 per million characters and focuses on higher-quality speech. The Flash version costs $15 per million characters and is designed for situations where lower delay and lower cost matter more.
MAI-Voice-2.1 supports 23 languages and 26 locales while keeping the same voice style between languages. That continues Microsoft's recent push to develop more AI models internally rather than relying entirely on outside providers.
Earlier MAI speech models have already been integrated into products such as Copilot and Teams. Microsoft has also made its models available to developers through Microsoft Foundry.
What Is the Price Target for Microsoft?
Turning to Wall Street, analysts have a Strong Buy consensus rating on MSFT stock based on 34 Buys and one Hold assigned in the past three months, as indicated by the graphic below. Furthermore, the average MSFT price target of $575.91 per share implies 11.8% upside potential. (See MSFT Stock Forecast).