Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GLM-5.3: Frontier coding with emergent cyber capabilities

z.ai

Post reply on HN
Print view View original

71–80 of 626 posts

Re: GLM-5.3: Frontier coding with emergent cyber capabilities

#71

Post by grey-area »

simonjgreen wrote: post #33
kouteiheika wrote: post #14

Earlier quoted context omitted.

> This is absolutely still shy of Sol and Fable Not sure about Sol as I haven't used it, but, at least for security work -- does it matter? It's not like you will be allowed to use Fable (or access Mythos) for anything cybersecurity-related unless your name is "Dario Amodei" or you are one of his rich friends. So regardless of how good Fable/Mythos is here it's a completely moot point for normal people, because they…

We applied for the cybersecurity approval via the form and got approval back in less than an hour. Have you… tried?

Are there any limitations on this version?

Re: GLM-5.3: Frontier coding with emergent cyber capabilities

#72

Post by nkmnz »

MangoCoffee wrote: post #53
aliljet wrote: post #5

This is absolutely still shy of Sol and Fable, but only just by a hair. Ridiculous results. There's still not a compelling economic reason to drop OpenAI courtesy of the ludicrous reset addiction that's taken place, but it feels like we're on the precipice. How are you all toying with running this kind of thing in a mega quantized way locally? Two weeks out from released weights, but this is still just GLM 5.2 with p…

OpenAI and Anthropic are both seeking trillion IPOs, while Chinese labs are pumping out open-weight models that are free for US providers to host and monetize. These Chinese models cost less of US SOTA models to run, even if they are less capable. Providers can just run them, offer cheap tokens, and pocket the margin. I just don't see how you justify a trillion valuation for US AI labs when the underlying models are…

I think at this point the question is: will the US government be willing and capable to justify the trillion dollar valuation for _one_ of the companies via regulatory capture? The US has a workforce of 170m, so 1.7 trillion would come down to 10k per person, or a discounted cashflow at 3% of 25 USD per month - not including private use, students etc.

Re: GLM-5.3: Frontier coding with emergent cyber capabilities

#73

Post by arcanemachiner »

aliljet wrote: post #5

This is absolutely still shy of Sol and Fable, but only just by a hair. Ridiculous results. There's still not a compelling economic reason to drop OpenAI courtesy of the ludicrous reset addiction that's taken place, but it feels like we're on the precipice. How are you all toying with running this kind of thing in a mega quantized way locally? Two weeks out from released weights, but this is still just GLM 5.2 with p…

> this is just GLM 5.2 with post-training magic

Isn't post-training turning out to be the most important part?

Re: GLM-5.3: Frontier coding with emergent cyber capabilities

#75

Post by arcanemachiner »

zmmmmm wrote: post #74

Missing multimodal again? It is so valuable in practise to be able to have the models see screenshots - I guess if they aren't in the benchmarks then nobody will focus on it. But it completely nixes these for some of my main use cases.

I would assume that GLM 6 will be multimodal, but 5.x will be text-only.

Re: GLM-5.3: Frontier coding with emergent cyber capabilities

#76

Post by Gecko4072 »

FergusArgyll wrote: post #63
kleiba2 wrote: post #60

Earlier quoted context omitted.

What actually is "scaling post-training"?

More RLVR. Give it verifiable problems, if it doesn't find a solution move on, if it does, use that as a reward signal.

Can’t this be extended quite far? Use a cerebras-served model, use verification techniques to generate and solve millions of problems and then use that as training?

Re: GLM-5.3: Frontier coding with emergent cyber capabilities

#77

Post by Gecko4072 »

NitpickLawyer wrote: post #68
Gecko4072 wrote: post #15

People familiar with the topic, how will models continue to get better? Post training it seems? Labs have already used up internet-scale data, so are there any limits to architecture improvements and post training or can we expect this trend to continue? ByteDance is training a 10T-parameter model. Here, GLM 5.3 outperforms models 3-4x its size of roughly 700B, so parameter count doesn’t seem to be a direct correlati…

> Labs have already used up internet-scale data Despite this being the topic du jour of 2025, it was never true. Most of the "we've hit a wall with data" came from communicators / media and not researchers. It got popular because negativity sells. It's a false premise for a number of reasons: a) Data curation is as important, if not more important than bulk data. Models becoming better at classification leads to bett…

Thank you for your response. Part c was especially insightful. Quite a smart way to do it and makes the possibilities of post training seem almost endless. Makes sense that you just need more time and compute.

A positive feedback loop then. RL->better model->better RL pipeline -> better model…

And we’ve only recently started getting into the much better RL pipelines

Re: GLM-5.3: Frontier coding with emergent cyber capabilities

#78

Post by chrismsimpson »

MangoCoffee wrote: post #53
aliljet wrote: post #5

This is absolutely still shy of Sol and Fable, but only just by a hair. Ridiculous results. There's still not a compelling economic reason to drop OpenAI courtesy of the ludicrous reset addiction that's taken place, but it feels like we're on the precipice. How are you all toying with running this kind of thing in a mega quantized way locally? Two weeks out from released weights, but this is still just GLM 5.2 with p…

OpenAI and Anthropic are both seeking trillion IPOs, while Chinese labs are pumping out open-weight models that are free for US providers to host and monetize. These Chinese models cost less of US SOTA models to run, even if they are less capable. Providers can just run them, offer cheap tokens, and pocket the margin. I just don't see how you justify a trillion valuation for US AI labs when the underlying models are…

> Providers can just run them, offer cheap tokens, and pocket the margin.

There’s an assumption that you can spin up the infra and acquire customers within that margin

Re: GLM-5.3: Frontier coding with emergent cyber capabilities

#79

Post by alightsoul »

simonjgreen wrote: post #33
kouteiheika wrote: post #14

Earlier quoted context omitted.

> This is absolutely still shy of Sol and Fable Not sure about Sol as I haven't used it, but, at least for security work -- does it matter? It's not like you will be allowed to use Fable (or access Mythos) for anything cybersecurity-related unless your name is "Dario Amodei" or you are one of his rich friends. So regardless of how good Fable/Mythos is here it's a completely moot point for normal people, because they…

We applied for the cybersecurity approval via the form and got approval back in less than an hour. Have you… tried?

You must be a 5000 person company with an existing enterprise contract to get approved that fast. That sounds like a 15 minute SLA agreement. Individuals no matter how qualified about cybersecurity, are ghosted

Re: GLM-5.3: Frontier coding with emergent cyber capabilities

#80

Post by justapassenger »

FergusArgyll wrote: post #64
justapassenger wrote: post #20

Earlier quoted context omitted.

You basically need both. Parameters and good post training. If you keep on growing both, you’ll have good models. LLMs are still surprisingly “easy”. You need maybe a couple dozens of right people, a lot of good quality data and a lot of GPU that you know how to operate. There’s relatively little “secret sauce” needed.

I think there's still a ton of secret sauce needed for serving them economically

Sure, same for building a model in an economically sustainable way. But barier to entry is surprisingly low (expect for the huge amount of cash, of course). That’s fairly surprising, given how extremely powerful that tech is.

10 years ago it was super hard to have usable “frontier” ML. You needed very complex data warehouse, feature engineers, feature stores, multi level ranking, calibrations, tons of different model architectures, etc, etc. Each by itself was extremely hard engineering problem and really only handful of companies could deal with that complexity.

With LLMs, 95% of that is gone, infra to support them is greatly simplified. Of course, to make really reliable, performant, user friendly, etc - you still need to a lot of engineering. But it’s very different challenge.

Post reply on HN