Meet Microsoft's in-house AI models that rival ChatGPT, DeepSeek and Gemini
Microsoft is taking a bold step toward AI independence with the release of two models developed entirely in-house. While its Copilot assistant still runs on OpenAI’s GPT tech, the new models signal Redmond’s intent to compete more directly with ChatGPT and Google’s Gemini. Here are the key takeaways:
Two new models unveiled Microsoft introduced MAI-Voice-1, its first natural speech generator, and MAI-1-preview, a text-based foundation model built from scratch.
2/5
Voice model in action MAI-Voice-1 is already powering Copilot’s Daily and Podcast features, delivering more human-like speech output.
3/5
Text model in public testing MAI-1-preview is available for testing on LMArena and will gradually be rolled out into Copilot integrations.
4/5
fficiency at the core Mustafa Suleyman, head of Microsoft AI, highlighted that MAI-Voice-1 runs on a single GPU while MAI-1-preview was trained on 15,000 Nvidia H100 GPUs—far fewer than the 100,000-plus reportedly used for xAI’s Grok.
Strategic independence While Microsoft remains deeply invested in OpenAI, developing in-house models shows its long-term plan to reduce reliance on partners and compete head-on with OpenAI, Google, and Anthropic.
Read the latest and trending tech news—stay updated on AI, gadgets, cybersecurity, software updates, smartphones, blockchain, space tech, and the future of innovation.
