Why China is releasing open models

One common theme we have seen in last few years is that whenever OpenAI or Anthropic or any private firm release a new AI model with crazy benchmarks, within a month if not days a new open weighted model gets launched which is almost closer to that benchmark. And this open weighted model is generally from some Chinese company Alibaba, Mooonshot {name more}. Latest one GLM-4.6 by Zhipu which is almost as close as opus 4.8.

The natural question comes in mind that why China is doing this. Aren't China a communist state where everything is controlled by the government, at least that is the perception of China in eyes of most of the world. Open source always saw contributions from Europe and America more in any of the previous tech revolutions starting from software for the personal computers to internet to even mobile software which even China used. But in this AI revolution why all of a sudden china is contributing towards worlds open source revolution?

Open source vs Open weight

Before going forward its important to state that there is distinction between open source and open weighted . most of the models which are being released by china are open weighted not open sourced.

What is open source in terms of AI?

We can call a model open source only when all the training parameters as well as what data its being trained on is available. anyone can reproduce or at least fork it to make the upgradation on it. but the models like glm,qwen, kimi and other are open weighted. only the team working on it can do the upgradation to there models.

so its the same as openai or anthropic's models? well no. most important difference is that open weighted models can be downloaded and can be used even without the internet, once it lives on one's machine. no data can go off your machine if you don't want. and its free to use. So yeah it is a big deal that one can use opus 4.8 level model free of cost if one can afford hardware required for it which costs a fortune in itself.

--not finished