Nvidia, Microsoft and IBM back open-weight AI models

Wo das Thema steht

The dispute over open-weight artificial intelligence has moved outside the United States. Moonshot AI, a Chinese start-up, has published the weights of Kimi K3, a model of 2.8 trillion parameters, for free download, making it the largest openly distributed system so far; benchmarks run by the company itself, and not independently verified, place it ahead of the strongest closed models from Anthropic and OpenAI, and American developers and firms are reported to be switching to it on cost grounds. That gives the position endorsed by Nvidia, Microsoft and IBM, which favour distributing model parameters rather than holding them closed, a concrete and commercially attractive example at the frontier. It also sharpens the objection raised by Anthropic and OpenAI, who according to the New York Times broke with much of the rest of the sector over whether Chinese open-source models should circulate freely in the United States, since the system now drawing Western users is precisely the kind they wanted restricted. The argument has at the same time escalated beyond commercial disagreement. Washington accuses Moonshot of building Kimi K3 on Anthropic's technology, an allegation Beijing calls groundless, and no evidence for it has been reported. Much still rests on claims rather than confirmed facts: the performance ranking comes from the developer, the provenance accusation is unproven, and the American open-weight camp has still announced no specific models, licences or joint commitments of its own. Nor has any rule, export measure or restriction followed from the split described by the New York Times. What matters next is whether independent testing bears out Moonshot's benchmark claims, whether the accusation about the model's origins hardens into a government action, and how far Kimi K3 spreads through Western infrastructure while the question remains open.

AI models are described as "open-weight" when the trained parameters that define their behaviour are made available for others to download, run and adapt, in contrast to closed models that can only be reached through a provider's own interface. The distinction matters for who can inspect, modify and deploy a model independently of the company that built it. Nvidia, Microsoft and IBM are all major suppliers in the AI market, spanning chips, cloud services and enterprise software, so their stated positions carry weight in how the industry settles the question. This topic tracks their reported endorsement of the open-weight approach and what follows from it. The reporting so far comes from The National.