The Deepseek announcement surprised some who believed China lagged behind in AI development, but it shouldn’t have. The Qwen model family from Alibaba, for example, has been demonstrating impressive capabilities for over a year, often matching or even exceeding Llama models in smaller sizes. It's no surprise they're now pushing into even larger model development.

What's truly remarkable is the speed at which Chinese researchers are matching major model releases from the US – often within just a month or two. They're also consistently publishing high-quality research papers, contributing significantly to the field. (For fun, take a look at the author names from the top daily AI research paper for each of the past several days, excluding anything related to Deepseek.. I think you'll be surprised.)

For months, the leading open-source video model has originated from China. It's clear that China is a major player in the AI race, and their contributions deserve greater recognition.

There are no borders around AI for the moment, especially because so much of it is Open Source, but we should all be acutely aware that just because something isn't in our immediate field of vision, doesn't mean it isn't real.

Sources:

* Timeline of Qwen development: https://lnkd.in/e9AY3SWF

* Benchmark comparison of Qwen and Llama from 4 months ago: https://lnkd.in/eHs3d2Tf

* Samples, github, project, report for Tencent's Hunyuan video model from early December: https://lnkd.in/eV9vArTK