Join the discussion
Write your take first — we'll ask for email only when you're ready to publish.
- Hacker News
- This is such a fundamentally misguided policy. The real competition isn't where the models are today. It's where the innovation will be tomorrow and what that should look like.
Trying to protect existing incumbents by banning Chinese open weight models is only going to stack the deck further in favour of Open AI and Anthropic. The real edge that the United States had was never excelling in the here and now. It was in being one step ahead in figuring out what the 'next big thing' was going to be. Propping up a duopoly that squeezes out all other competitors is the very opposite of that.
Software and technology is probably the last bastion where the United States still has somewhat of an ability to out-compete the Chinese. Our lead in this area came from allowing upstarts and innovators to kill off and replace the fossilized incumbents, not by propping up the big shots.
I am genuinely bewildered as to why we would be hell-bent on destroying that advantage.
by khriss - The point of open weight models is that this is not possible, companies will be built by taking 'contraband' chinese open weight models and marketing as something from the USby jgbuddy
- Without a massive amount of post-training, it will be obvious that they are Chinese models with Chinese political ideology. I doubt that there's any business model that would work there.by intrasight
- A "ban" is a legal prohibition. The possibility of the ban is solely a legal decision. Open weights might make it easy to violate a ban, but it does absolutely nothing to prevent one from being possible. The questions as to whether a ban would be effective hinges entirely on the ability and desire to enforce it.by kube-system
- You don't need an archive of huggingface. Instead you can use the Chinese version of huggingface with all Chinese models - https://www.modelscope.ai/modelsby shixinhb
- Blocking Chinese models would just be regulatory capture by another name. If we want OpenAI and Anthropic to improve they need to compete not just with eachother but with foreign labs too.
I don’t know why anyone would think giving Mango Mussolini the power to decide what AI models they can use is a good idea.
- It's not really a competition with the foreign labs cause they're copying. To illustrate the difference, if there were somehow an impenetrable firewall between the US and China, the US models would continue improving while the Chinese ones don't.
The only solution is for US labs to find a way to not get copied, and idk how. Banning Chinese models doesn't help with that.
by frollogaston - Could foreign labs compete with openAI and anthropic if they weren't literally just copying their outputs?
Can the foreign labs actually advance the state of the art if they put openAI/anthropic out of business?
- But the idea of a blanket ban on Chinese open-weight models was not seriously discussed.
Article should probably put the above higher in the story.
- As if making something impossibly difficult to comply with is not defact ban. Case in point: OFAC sanctions waivers. Show me a bank in US that directly deals with those and you will quickly end up with rather interesting representation that mimics current AI situation.
- Dean W Ball from OpenAI kinda laid down how the “ban” would work. Everyone clowned him over the weekend, but pretty much everything he said had substance. It’ll most likely will go that way - “mention in memos in all gov orgs they’re not supposed to use any AI other than American ones. And let the implications handle everything else, rather than legalese”by tokioyoyo
- If there were ever a sign I need to get mirroring, this is it.
What are the best Chinese models on HuggingFace today? Bucket by ideal RAM: <16GB, <32GB, <96, <256, 256+
Text generation, image generation, TTS, etc
by 3eb7988a1663 - qwen3.6-27B, gemma4-it-31B, poolside/Laguna-S-2.1 are all good text gen models generally in the <128GB (depending on quant, you can fit them in smaller mem footprint)
Note, Laguna provided quants have some issues and they are reworking/updating them.
by verdverm - It is a lost opportunity that huggingface (or unsloth) frontpage doesn't have a list like this that updates which each model release.by throwa356262
- Specific use case is way too important for a list like that to be very helpful. I only know image gen well enough to use as an example, but style, resolution, inpainting, img2img or text2image, lora compatibility, etc. all lead to different “best” models.
Even more complex for image2video, but there’s fewer models to choose from there at least.
by Lord-Jobo - Hugging Face recently added a filter directly on the models page: https://huggingface.co/models that lets you filter based on hardware.by abidlabs
- My list would be to go grab GLM5.2, Deepseek v4 Flash and Pro, and the various sizes of Qwen 3.5/3.6 dense and MoE, along with at least one of the multi modal qwens (I think it’s called VL). In a pinch, you can quantize them yourself to whatever size you need.by ericd
- The two major image generation models are both by Alibaba - Z-Image and Qwen-Image-2512/Qwen-Edit-2511. They are both decent but have different use-cases. Unfortunately with the recent Qwen-Image 2 and now 3 it doesn't seem likely that they're going to be doing any more open-weight image gen.
https://huggingface.co/Tongyi-MAI/Z-Image-Turbo
by vunderba - Alibaba Cloud runs ModelScope, which is like Hugging Face but based in China:
- Source code (ironically on GitHub): https://github.com/modelscope/modelscope
It's not limited to Alibaba's Qwen. All of the other major Chinese ones are on there (GLM, DeepSeek, Kimi, etc.) as well as finetunes and quantizations of the non-Chinese ones.
by commoner - 256+: GLM-5.2 (swap with Kimi K3 when it goes open-weight on Monday)
<256: Actually, surprisingly, not a Chinese model but probably Laguna S2.1. The best Chinese model at this size is DeepSeek V4 Flash though
<96: Qwen 3.6 27B
<32: Still Qwen 3.6 27B (NVFP4)
<16: Oof, not sure. Nothing will feel great at this size TBQH without finetuning on a specific task. Pick your poison of tiny Qwen or tiny Gemma (although again Gemma is not Chinese)
by reissbaker - You guys need to wake up and realize that Anthropic are the enemy. They just increased their political spending as well... Stop giving them your money or this never ends.by nullbio
- Can you elaborate on what makes you say this? I'm really struggling to understand the source of all the Anthropic hate on HN recently.by vaaal
- Someone explain to me how the US would stop people from using Chinese models?
Anyone in Europe can download and run a Chinese model and serve it up on the open internet to people in the US. What can the US do about that? The only thing the US can maybe do is ban exports of high powered GPUs to the EU. However the EU can retaliate by banning export of ASML machines to the US, so I don't think the US has that card to play.
by zarzavat - The US will stop you from using it with sanctions. As a US citizen anywhere in the world intentionally circumventing sanctions grant you up to 20 years prison time.
That’s just the short version not to speak of secondary sanctions and further mechanisms which take place then.
by sharpshadow - The US is trying to stop them for US companies, it doesn't care about non Americans.by satvikpendem
- Companies would simply move away from the US. Kind of like how Britain killed it's own car industry back in the day.by dosisking
- The government can strong arm enterprises in a million ways. If they just issue a vague executive order large public companies won't want to buy inference from these models without regulatory certainty. Anyone in regulated industries is going to be weary of regulatory retaliation.
I.e. for Google to build data centers they need the US gov to play ball, so if they were considering a service of hosting open source models on their TPUs they wouldn't do this. Another example is the merger between Paramount and Skydance where they paid out Trump to get the merger approved [1].
[1]. https://www.yahoo.com/news/paramount-settles-donald-trump-la...
by ianm218 - Probably in a similar way to how they go after IP/copyright currently. One method they use is to push for law changes in free trade agreements. US would use whatever leverage they have to get other countries on board and locked in to US providers.by dmuso
- Just declare it illegal (national security risks). Treat it like sexual abuse material, scan cloud services for copies, and jail everyone who has them.by andix
- Read this thread when you get the time:
https://x.com/_vkaku/status/2080352797606744209
Open Data+Open Models gives everyone else an advantage and bringing regulatory capture here should be appealed and brought to the FTC and the courts to challenge such regulations.
Startups need better than this whole lock down into four overvalued frontier models in the US sort of thing
by vkaku