China makes open-source large models work longer and become smarter

China has achieved a breakthrough in open-source large language models, enabling them to work longer and become smarter. The development focuses on enhancing the reasoning capabilities and extended operational duration of open-source AI models.

Moonshot AI

Following DeepSeek's architecture, US big models start copying China's homework

The article reports that American AI companies, including Meta, are adopting architecture ideas from China's DeepSeek model. This indicates a shift in the AI development landscape where US models are increasingly inspired by Chinese innovations.

DeepSeek Meta