Microsoft's AI chief told staff the company quietly trained a model that outperforms a key partner's flagship system at a ...

Microsoft has completed pre-training a new AI model with 2.8 trillion total parameters, saying it outperformed every open-weight model it tested while requiring up to 12 times less computing power to ...

Base-2 packs 2.8 trillion parameters and trains 4 to 12 times more efficiently than DeepSeek, but public benchmarks are ...

Microsoft has announced that it has completed pre-training of the AI model "MAI-Base-2," which has a total of 2.8 trillion ...

New plans for Azure could leave less room for Nvidia.

Base-2, a 2.8-trillion-parameter vision-language model. We separate the loss and compute denominators behind the "up to 12x" ...

Can in-house chips pop the AI bubble? Microsoft's Maia 300 makes the case, but Nvidia shrugged off the news with a 1.29% ...

Compare Microsoft Maia 300 vs Nvidia Rubin NVL72 on HBM4 memory, bandwidth, rack design, availability, and efficiency claims ...

The Microsoft Surface Laptop Ultra isn't just a MacBook Pro for PCs. It's a strong sign that local compute power is what we ...

Microsoft announced yesterday that it has completed the pre-training of its next-generation SOTA foundational model, MAI-Base ...

Microsoft has linked two Fairwater data centers 700 miles apart into one AI superfactory, cutting model training time from months to weeks.

Microsoft announced yesterday that its next-generation SOTA foundation model, MAI-Base-2, has completed pre-training and ...