I am saying this probably is "silly behavior by a government" and it is a milestone that points towards what the future may look like. Why can't it be both?

It's easy to wave this aside as the current administration playing political games. But I don't think there is any reason to assume that the current era of open availability of models is going to continue indefinitely. Do you think that Chinese labs will continue to release open models forever, even why they get to the level that Mythos is at now, and beyond? And do you think that a competent US government would have no interest in regulating and restricting model access in 2 years time, assuming that model capabilities continue to improve? I think we bias towards thinking the status quo is the norm and will continue, but this news invites us to question that assumption and think about different ways the future could go.

▲

gpm 3 hours ago | parent | next [-]

> Do you think that Chinese labs will continue to release open models forever

Yes.

I think the Chinese government either already has, or will soon, grasp that if they train the models that people use they dictate what people believe (at least around the margins where that's malleable), and they will happily throw resources at that.

And simultaneously that the only way they can actually get everyone to use their models is if it's possible for us to run them on our own hardware.

(This isn't exactly a utopian view of the future)

▲

jychang 2 hours ago | parent | next [-]

This is going to age very poorly when the best Chinese labs ALREADY just started not open sourcing their models.

Qwen 3.7 is not open source; previous Qwen versions would have open source releases, but Qwen 3.7 plus does not. The second best Chinese model, Minimax M3, is testing the waters by taking longer and longer between “model release” and open sourcing it. This time, they spent 2 weeks after release before open sourcing it. There’s also a lot of rumors of GLM and Deepseek not open sourcing future models.

It’s pretty obvious that you cannot take Chinese models as open source for granted, they’ll be closed source soon.

▲

ls612 2 hours ago | parent | next [-]

The main reason the Chinese labs are releasing models as open weights is because they don't have the compute necessary to provide all of the inference. For the US frontier models something like 80-90% of the lifetime compute required for the model is inference rather than training. China wants to shepherd as much of their limited compute as possible towards training to keep up in the race.

	▲	londons_explore an hour ago \| parent [-]
		With nearly everyone using inference accelerators, the pool of hardware is no longer shared between training and use.

▲

zardinality an hour ago | parent | prev [-]

[dead]

▲

nine_k 2 hours ago | parent | prev | next [-]

The US administration restricting the use of US-trained models is one of the best gifts it could make to the Chinese LLM producers, and to the PRC government.

▲

dozerly 2 hours ago | parent [-]

This entire administration is a gift to everybody but the US. It’s either in service of Russia, China or whoever is willing to pay Trump the most.

	▲	rjzzleep an hour ago \| parent [-]
		Chinese have a nickname for Trump. 川建国. Trump the nation builder(meaning China). But Biden actually continued most of Trumps policies.

▲

tw1984 3 hours ago | parent | prev [-]

> I think the Chinese government either already has, or will soon, grasp that if they train the models that people use they dictate what people believe (at least around the margins where that's malleable), and they will happily throw resources at that.

that doesn't require the model to be SOTA, it can be just a compact model capable of running on some inexpensive hardware. that is vastly different from SOTA models like Mythos which can potentially disrupt lots of things.

▲

strangegecko 2 hours ago | parent [-]

Of course it requires SOTA, people will always choose better models over some compact thing that is obviously more limited. You can't control the truth with models nobody wants to use.

▲

columnarx3 an hour ago | parent | next [-]

People choose SOTA right now because of the heavily subsidised model subscriptions. People aren't going to pay 20x the price for a model that's maybe 10% better.

	▲	ezst 39 minutes ago \| parent [-]
		And the fact that "better" is highly subjective and domain/task/vibe-specific

▲

adrianN an hour ago | parent | prev | next [-]

Why do I want the model I use for coding to know Shakespeare or vice versa?

▲

rjzzleep an hour ago | parent | prev [-]

Small models are the future.

▲

deanishe 2 hours ago | parent | prev | next [-]

> Why can't it be both?

Is the government going to fund all further development? Hard to imagine investors continuing to throw billions at products they aren't allowed to sell.

▲

locknitpicker 2 hours ago | parent | prev [-]

> I am saying this probably is "silly behavior by a government" and it is a milestone that points towards what the future may look like. Why can't it be both?

Here is why it's unlikely this is anything other than "silly behavior by a government":

- some benchmarks show GPT-5.5, Gemini 3.1, and even Claude Opus outperforming Claude Fable, and yet it's Fable which is restricted.

- some benchmarks still show the likes of Kimi 2.5 outperforming any Claude model, and DeepSeek is getting equivalent scores (a few tenths of a percent difference)

> Do you think that Chinese labs will continue to release open models forever (...)

That's immaterial to the discussion. Even if China forced Chinese labs to restrict access to all models, the truth of the matter is that Trump's administration to restrict access to US-based models does not prevent others from having access to models that are as capable or even better.

So what's exactly the point of this?

▲

rileyphone 2 hours ago | parent | next [-]

All that says is some benchmarks aren’t worth the tokens it takes to evaluate them. Mythos is clearly capable of finding zero days other models can’t, and Fable is close enough to be lumped with it.

	▲	mullingitover 33 minutes ago \| parent [-]
		> Mythos is clearly capable of finding zero days other models can’t I'm unconvinced that this is anything more than proof of work and marginal improvement that other models will catch up with, perhaps as early as to next week. Lots of other current-gen models will find vulns that can be chained together if you're willing to burn enough tokens on the task, and Fable is an absolute token incinerator.

▲

solumunus 2 hours ago | parent | prev | next [-]

You’re completely overrating these benchmarks and it’s landing you at a nonsense opinion. Just actually use the models and you will see that the gap is significant.

▲

kolinko an hour ago | parent | prev [-]

Did you use the models yourself?