I swear, we do this every 3 months. "Chinese models are falling hopelessly behind. They don't have enough compute". And then they release their new models "Oh, they are caught up. Need to ban them".

What happened: A post argues that claims about Chinese AI models falling behind alternate with claims that they have caught up and should be banned, and says the gap looks smaller when flagship models are released. Key numbers: The post says this cycle repeats every 3 months and that updating every 3 months is not great. Why it matters: The post says comparing frequent release models, such as DeepSeek Flash versus Sol, is a better proxy for how US and Chinese labs are progressing, and suggests kimi and GLM should update faster.

A post argues that claims about Chinese AI models falling behind alternate with claims that they have caught up and should be banned, and says the gap looks smaller when flagship models are released.

The post says comparing frequent release models, such as DeepSeek Flash versus Sol, is a better proxy for how US and Chinese labs are progressing, and suggests kimi and GLM should update faster.