Microsoft makes “significant investments” in training of own AI models

Microsoft AI has launched its The first internal model Last month, adding an already complicated relationship with their Open’s partner. Now, Microsoft and Head Mustafa Suleiman says the company brings “significant investments” in computer capacity needed for Microsoft’s own construction models.

“We should have a capacity to build a world-class border models in the house of all sizes, but we should be very pragmatic and use other models where we need,” Suleiman said during Microsoft City Castles in Microsoft City Hall on Thursday. “We will also bring significant investments in your own cluster, so today Mai-1-review is dressed only at 15,000 H100, a tiny cluster in a big scheme of things.”

Suleiman indicated that Microsoft has ambitions to train models that are comparable to the efforts of targets, Google and XAIs on clusters who have used six to ten times, which Microsoft used for your Mai-1-review. “More to do, but it’s good to take the first steps,” Suleiman said.

Microsoft CEO of Satia Nadella said in the same city councilor to “look forward to the possibility of building a model, so we can build products forward forward.” Nadella also clearly gave Microsoft “definitely support multiple models” in its products and prominent Github Copylot as “the best example” of that strategy.

Microsoft, allegedly planning to use anthropic and models for some functions in Microsoft 365 shortly. Info It reports that Microsoft 365 Copylot will partially launch anthropic models “, after Microsoft revealed that some of these models surpassed Openai in Excel and PowerPoint.