Microsoft has released three AI models created in-house. This is a move that signals increased independence from long-time partner OpenAI.
MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 are designed for transcription, audio generation, and image creation and are available exclusively on Microsoft’s Foundry platform, the company announced in a blog post Thursday.
Microsoft Foundry is a platform within Microsoft Azure that enables developers and enterprises to buy, build, and customize AI applications using a variety of models.
This model is available to enterprise customers and will therefore be in direct competition with similar tools from OpenAI.
OpenAI offers Whisper for transcription, Text-to-Speech models for audio, and DALL·E for image generation. All three are also available at Foundry, either as individual products or as part of OpenAI’s broader model.
Microsoft AI CEO Mustafa Suleiman said in a post on X on Thursday that MAI-Transcribe-1 is “the world’s most accurate transcription model” and MAI-Voice-1 “sets a new standard for natural speech.”
Microsoft is OpenAI’s largest investor and the two companies remain close partners. OpenAI relies on Microsoft’s Azure cloud services to train and run its models, and Microsoft uses OpenAI’s ChatGPT to power its AI chatbot, Copilot.
But Microsoft is increasingly asserting itself in the AI race, reducing its reliance on OpenAI by building its own frontier models.
In November, Suleiman assembled a superintelligence team at Microsoft to focus on training frontier models at all scales with its own data and state-of-the-art computing, making the company “self-sufficient in AI,” he told Business Insider at the time.
Previous agreements between the two companies determined how Microsoft could independently develop advanced AI systems. a New contract announced in October This allows companies to “pursue AGI (artificial general intelligence) independently or in collaboration with third parties.”