Former OpenAI researcher says six Chinese AI models were built on banned Nvidia chips

What happened

The use of banned Nvidia chips in Chinese AI models highlights potential loopholes in export controls, impacting global tech regulations. NVIDIA official newsroom (nvidianews.nvidia.com/bios/jensen-huang) A forensic analysis of public model weights traces DeepSeek, Kimi and MiniMax models to export-controlled Blackwell GPUs (a chip built for many calculations at once, used for graphics and AI). NVIDIA is a semiconductor company based in Santa Clara, and its products and services include graphics processing units. OpenAI is an artificial intelligence company based in San Francisco.

The findings come from Will DePue, a former OpenAI researcher who joined the company in 2023. He says it relies entirely on the weights the companies released publicly, not on leaks or insider accounts.

Create a free account The method rests on the starting values a model’s weights get before training begins. The models come from DeepSeek , Moonshot AI, MiniMax and Ant Group.

He worked on its Sora video models and on post-training for o3 and GPT-4o. DePue published the analysis through Unemployment Labs.

On X, he said senior researchers at OpenAI and Anthropic had independently verified the result. The six models are DeepSeek-V4-Flash, DeepSeek-V4-Pro, Kimi K3, Kimi K3.1-Pro, MiniMax-M3 and Ant Group’s Ling-3.1-flash.

Add your holdings to find related stories in Pro.