Learn extra at:
Independence day
In response to Microsoft, MAI-1-preview makes use of an in-house mixture-of-experts mannequin that was pre-trained and post-trained on 15,000 Nvidia H100 GPUs, a extra modest infrastructure than the 100,000 H100 cluster sizes reportedly used for mannequin improvement by some rivals. Nonetheless, with a watch to ramping up efficiency, Microsoft AI is now operating MAI-1-preview on Nvidia’s extra highly effective GB200 cluster, the corporate mentioned.
For Microsoft, and maybe your entire AI sector, the fashions’ look marks an essential second. Hitherto, the corporate has relied on fashions from OpenAI, an organization it has invested a reported $13 billion in since 2019 and which makes use of the Azure cloud platform to host its fashions and companies.
Nonetheless, in 2024 Microsoft stunned the trade by hiring former DeepMind founder Mustafa Suleyman and his Inflection AI startup group to develop AI fashions for Microsoft AI, unbiased of OpenAI. Inflection AI continues to develop its personal enterprise AI with a brand new CEO, whereas the previous founders and core group work on Microsoft’s Copilot merchandise.