Microsoft Introduces New AI Models, Signalling Shift from OpenAI Reliance

Prime Highlights

Microsoft launched MAI-Voice-1 and MAI-1 Preview, marking a direct challenge to OpenAI and other AI leaders while reducing reliance on outside models.
The company envisions building a continuum of specialized AI tools, reinforcing its long-term strategy in the rapidly growing AI sector.

Key Facts

MAI-Voice-1 can generate one minute of audio in under a second using just a single GPU, now powering Copilot Daily, Podcast features, and Copilot Labs.
MAI-1 Preview, trained on 15,000 Nvidia H100 GPUs, is ranked 13th on the LMArena leaderboard, competing with GPT-5, Gemini 2.5 Pro, Grok-3, and DeepSeek R-1.

Background

Microsoft has announced two new AI models, MAI-Voice-1 and MAI-1 Preview, openly competing with OpenAI and other worldwide AI companies. Until now, Microsoft’s Copilot features have utilized OpenAI’s GPT technology. With this new launch, Microsoft communicates its push to be in the AI business, building its own systems while decreasing its outside dependency.

MAI-Voice-1 is open to natural speech generation and is characterized as one of the most effective models in the market. Microsoft claims it can make a minute of audio in less than a second, and it only requires a single GPU. Copilot Daily, Podcast features, and Copilot Labs are now powered by the model.

Microsoft also published MAI-1 Preview, a base model that can currently be tested on LMArena, a public large language model benchmarking platform.
The model is now ranked 13th on the leaderboard, with competitors such as GPT-5, Gemini 2.5 Pro, Grok-3, and DeepSeek R-1 remaining at the top.

Unlike rivals, which often use over 100,000 GPUs for training, Microsoft built MAI-1 on 15,000 NVIDIA H100 GPUs. AI Chief Mustafa Suleyman said the company adopted strategies from the open-source community to stretch resources further and focus on selecting the right training data.

Looking ahead, Microsoft is working on its next generation of models using some of the world’s largest data centers equipped with Nvidia’s new GB200 chips.

The company claimed that its vision is not one big model but a continuum of specialized AI tools for various applications. Microsoft projected a long-term goal in the fast-growing field of AI, and in a blog post, it stated that there will be much more to come.