Chinese AI critics stopped calling it stolen. Now they're arguing about the price.

For the first time in years, the leading Chinese-language objection to a top Chinese AI model has nothing to do with distillation. It has to do with price.

Moonshot AI's Kimi K3 sits at the center of the conversation. Per a Chinese feature outlet translated by Jeffrey Ding's ChinAI newsletter, K3's blended API rate is $2.30 per million tokens, ahead of Alibaba's Qwen3.7 Max at $1.40, more than twice Zhipu's GLM-5.2 at $0.90, and roughly thirteen times the rate for DeepSeek's V4. The same translated report tagged the positioning as "affordable luxury," a luxury import priced like one even when its domestic rivals sit closer to commodity territory.

For most of the last two years, the default Chinese AI headline in U.S. coverage has been defensive: Chinese frontier models were distilling American ones, the rankings did not mean what they seemed. When K3 dropped in late July, that script briefly reappeared in U.S. social media, in posts like "let's just ban it" and "this is only because they distilled American models," and then it did not survive contact with the Chinese-language response. There, the conversation pivoted to whether Moonshot was pricing itself into a corner.

K3 is, by several independent measures, the strongest open-weights model to come out of China. Per the same translated report, it carries 2.8 trillion parameters, sits third on the Artificial Analysis Intelligence Index behind what the source names as Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol, and tops Code Arena. The numbers were real enough that even Elon Musk posted a one-word verdict: "Impressive."

On the model, Chinese commentary has been largely positive. The dispute is on the price. The Chinese AI stack has trained its buyers, enterprise and consumer alike, to expect sub-dollar token rates, and DeepSeek's V4 in particular has anchored that expectation. When Moonshot set K3's blended rate at $2.30, Chinese social media split between admiration at the technical leap and a sharper question: who is this for, and why should a Chinese buyer pay a premium over a domestic rival when the open-weights story is that you do not have to?

The defensive posture survives, but it no longer carries the conversation. Chinese outlets still describe K3 as a "matter of honor for Chinese technology," and the distillation debate is unresolved. The U.S. line reappears whenever a new Chinese model lands, and the responses have not moved it. The closed-source tension is real too: K3 is open-weights but not fully open, and the gaps between those two categories are where the most uncomfortable questions sit. The new lead is the price.

Open-weights frontier models have to be paid for somehow, and U.S. frontier labs have shown that buyers will absorb ten- and twenty-times premiums when they believe a model is the best in class. By pricing above the Chinese commodity tier but well below the closed-source U.S. leaders, Moonshot is testing whether "best open model" can carry a "best Chinese model" premium without the closed-source label. The bet is that a slice of the Chinese market, and a slice of the global open-weights market, will pay a delta for the strongest domestic option, the same way some U.S. buyers pay a delta for Anthropic over a cheaper open alternative.

The next test is whether other Chinese frontier labs follow Moonshot up the price ladder or undercut it on the way down.