After years of relying on OpenAI’s technology in a turbulent partnership, Microsoft has finally introduced its first fully in-house AI models – MAI-Voice-1 and MAI-1-preview. With this move, the company signals its intention to take greater control over its own AI development and create models under its own brand.
MAI-Voice-1 is a speech generation model capable of producing one minute of speech in less than a second and is already integrated into apps such as Copilot Daily and Podcasts. On the other hand, MAI-1-preview is a text model specialized for instructions and everyday queries, trained on significantly fewer GPUs compared to its competitors.
CEO Mustafa Suleyman described MAI-1 as a model that is “among the best in the world,” although official benchmark tests have yet to be released. The text version is currently being tested on the LM Arena platform and via API, with Microsoft planning to gradually roll it out into specific use cases soon.
This shift marks a new phase in Microsoft’s relationship with OpenAI. Instead of being perceived as the “assistant” of another tech giant, Microsoft is now building its own AI infrastructure, enabling it to shape its future in artificial intelligence independently. The final verdict on the power of these new models will come from practice and independent evaluations, but it’s clear that the company is determined to prove it can set the pace in the AI industry.
In brief: Tech World Highlights
- Chinese companies are moving away from Nvidia’s H20 chips and turning to domestic solutions after being offended by remarks from U.S. Commerce Secretary Howard Lutnick.
- California-based FieldAI raised $405 million to develop foundational embodied AI models that enable humanoids and autonomous vehicles to operate in unfamiliar environments.
- The U.S. Navy has encountered setbacks in its swarm autonomous surface vessel program, Reuters reports, after two test incidents off the California coast, with one drone ship disabled.
- Figure AI released a video of its humanoid Figure 02 navigating obstacles using a new locomotion system called the Helix walking controller.
- Foxconn will reportedly unveil a humanoid with an “LLM-powered brain” later this year, to be deployed in a new Houston factory producing NVIDIA GB300 servers.
AI Trending Tools:
- Nemotron Nano 2 – Nvidia’s family of small, efficient reasoning models.
- Qwen-Image-Edit – Qwen’s new model for image editing.
- Eleven Music API – Integration of high-quality music into products and workflows.
