Jensen Huang calls Chinese models such as Kimi excellent, backs open source for AI efficiency

NVIDIA founder and CEO Jensen Huang has publicly weighed in on Chinese AI models, saying that models such as Kimi perform well enough, and voicing optimism that open source will help improve AI efficiency. The remarks were carried in a report by Sina.

Huang co-founded NVIDIA and leads the company whose GPUs sit at the center of AI training and inference. Because of that position, his public read on the model landscape is widely treated as a signal for where the industry is heading.

Kimi is the model family from Moonshot AI and one of the more closely watched offerings in China's model ecosystem. By naming it among the Chinese models he considers strong, Huang offered a direct endorsement of their capabilities.

The other emphasis in his remarks was open source. Huang said he expects open source to drive gains in AI efficiency, a view that tracks with a broader trend: open models are closing the performance gap with proprietary ones while often being cheaper to deploy and call.

Endorsement from the head of the industry's main compute supplier carries weight beyond technical merit. Stronger, more efficient open models can shift where inference demand lands and reshape how compute and models divide labor across the ecosystem.

What to watch next is whether the remarks are followed by concrete progress on open models and compute, and how Chinese models advance on efficiency and capability from here.