Feels like Alibaba's tokenplan is both expensive and limited, and the domestic models are just stupid

Alibaba’s token plan is expensive and has a tiny quota.

I saw their email this morning, clicked through, and was able to buy it. I got the 39-yuan quarterly package and connected it to Hermes, using the 3.8 Max model. The quota turned out to be ridiculously small—I barely asked it a few questions before hitting the weekly limit. I totally got scammed, and they don’t even allow refunds. TMD.

And, very importantly, the Qwen series models are damn verbose. I connected Hermes to WeChat, and WeChat’s iLink protocol has limits: sending too many messages triggers rate limiting. Then, for no apparent reason, Hermes often stops replying. Earlier, when I used a Codex subscription, I had this problem at first too, but after I reported it once, it never happened again. I assume those rules were added to its memory, but the Qwen model didn’t even look at them. It’s so damn stupid—a domestic model.


While you’re talking to it, it triggers iLink’s rate limit, and then Hermes stops replying. If you send it another message, it forgets the context. I never encountered this kind of situation when using Codex OAuth before. All I can say is that domestic models are incredibly stupid.

1 Like

Domestic models still need to be more multimodal—only Qianwen is passable so far.

2 Likes


As shown in the image, using 8M tokens with 80% cached has already consumed 68% of the weekly quota. If we estimate, for this whole 39-token plan, it’s probably only about 15M tokens. That’s way too little. I use all kinds of models—started with 3.8 max, then later switched to 3.7 plus, also used ds v4 flash and generated images with its wan image. Basically, I let Hermes auto-write articles and then just tweak the configuration a bit.

Final comparison: Aliyun’s token plan, which costs 39 RMB per month, has a weekly limit of only 18M tokens. It’s honestly more expensive than gold.

Wait until DeepSeek drives the prices down.

Waiting for Liangzi to post the pro version.

It’s really frustrating. Qwen is incredibly verbose—ask it just a few questions and it spits out a huge wall of text, burning through your quota in no time.

Yeah, the quota is really low. Honestly, I feel like it’s not even as good as just paying for got’s Plus—at least that way you don’t have to worry about quota issues all the time.

It’s true, Qwen’s replies are really wordy—it’s always a huge block of text, and your quota runs out really fast. Honestly, it doesn’t feel like a great value.