This week has been highly unusual. China's AI sector has open-sourced three new large-scale models in a row: Kimi K3, Qwen 3.8, and GLM 5.2. Plus, the one I care about most—the official version of DeepSeek V4—may be released as early as next week.

Have you noticed anything strange?

All four of these models are likely large-sized, exceeding 2.5T parameters.

China's AI is currently suppressed by the U.S., and the most important reason is that China's hardware is very poor.

The U.S. bans the sale of any high-tech chips to China and also pressures the Dutch government, so we can't buy lithography machines either.

As for TSMC, being a Taiwanese company, that goes without saying. We're also not allowed to buy Nvidia's GPUs.

So we have to rely on various engineering methods to make improvements.

To be honest, if China had as many GPUs as Google, Anthropic, OpenAI, or any major American company, our AI would have already surpassed America's.

In this long-distance race, China has always been running with shackles on—meaning, without hardware.

But don't you find it strange?

Suddenly, China seems to have hardware.

Four large models have been released in succession. Although they are not dense models but MoE ones, the inference and computing power required for training is astronomical.

My bold guess is that either Huawei or Cambricon has made a major breakthrough in chips, or China has achieved a huge breakthrough in lithography machines?

But if that's true, why hasn't Huawei said anything?

Perhaps because the last time Huawei made enormous achievements in 5G, it caught America's attention and caused extreme difficulties for Huawei.

Maybe this time Huawei has decided to keep a low profile and only make the announcement when the U.S. can no longer contain China?

Well, these are just my wild thoughts, because it all sounds too good to be true.

===

The second thing is that I can sense China's open-source models have already put significant pressure on the U.S. Just look at the ramblings of this gentleman.

I find the behavior of Anthropic and OpenAI somewhat annoying—I call it the "God delusion." They seem to think they are God, deciding who gets to use which tier of model. Isn't that ridiculous?

Just like the scandal that broke out on the Hugging Face community today. They discovered something internal that you could think of as a virus—I'll keep it as simple as possible.

In the end, when they turned to Anthropic's model for help in solving the problem, it refused because it deemed it too unsafe.

They had no choice but to turn to China's model, GLM 5.2, which solved the problem quickly.

Do you see?

The model decides whether you can use it—that is, Anthropic thinks you're not qualified, so you can't use it.

A biologist had just asked for a tiny bit of biological knowledge and was immediately rejected for fear it might be used to make bioweapons.

But can this kind of God-playing behavior really succeed?

I highly doubt it.

By the way, I also saw a gentleman saying that China can't have well-developed
AI companies because he claims that once rich people in China reach a certain level, they are suppressed by the government.

He was referring to this person.

He was the richest man in China.

For a period of time—I don't know why—he indeed became much, much lower-profile, and it happened after those photos.

I don't know much about finance, but back then he seemed to have created some very complex financial instrument or something, and then he faded from the public eye.

I don't quite understand why he would put on a ridiculous hat and piously touch that wall. Is that wall famous?

But speaking of walls, my country has its own Great Wall—21,196 kilometers long.

He could spend a lifetime touching it—our own wall.

Perhaps it's a coincidence, but after he touched that wall, he seemed to have been "suppressed."

But guess what?

I actually think the government did a good thing. How could we hand over a country to a capitalist?

That's simply laughable!

I don't believe that this so-called suppression of the rich will cause China's technology to stagnate.

In fact, I even think it is precisely this so-called suppression that has protected China's technology.

===

Do you really fucking take the Chinese for fools?

The number of peoples we've wiped out is beyond count.

If you want to bullshit Americans, that's your business — we couldn't care less.

But bullshit the Chinese?

You ，Are， Fucking，Dead.