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

GLM 5.2 Is Out

twitter.com

Post reply on HN
Print view View original

331–340 of 544 posts

Re: GLM 5.2 Is Out

#331

Post by vidarh »

naklitechie wrote: post #281
easygenes wrote: post #75

Announcement from the founder of Z.ai: “ GLM-5.2 is Fully Open, Frontier Intelligence Belongs to Everyone Today, the sudden restriction of certain frontier models is deeply regrettable. At a time when access to frontier models is abruptly cut off for non-technical reasons, we are even more convinced of one thing: science should be global. The path to AGI (Artificial General Intelligence) must never be enclosed by hig…

Looks like it's about a year behind. Not that I am complaining. A year behind is good progress. I also feel much of the trick is in the reasoning and harness. so some progress around that would accelerate this process.

Harness certainly matters a lot, though GLM is pretty forgiving. I just had Opus tell me that based on numbers over the last week, from quite a few billion tokens total across half a dozen providers, GLM 5.1 has been more reliable for one of my projects than Sonnet... Just switching on 5.2 now.

Re: GLM 5.2 Is Out

#332

Post by ThouYS »

ignoramous wrote: post #299
LaurensBER wrote: post #103

Earlier quoted context omitted.

Based on my first impressions it's about 6 months behind the frontier labs. So very similar to Opus in January. That is, pretty damn impressive and very useable. When it comes to architecture or complex problems it does noticeable worse but I don't think anyone expected anything else. One particular interesting strong point seems to be design and user interfaces. It does seem to punch above it's weight there but that…

> Based on my first impressions it's about 6 months behind the frontier labs. So very similar to Opus in January. According to this one benchmark, I find it amusing that Qwen3.6 27B beats ALL "frontier lab" models on coding Kotlin: https://archive.vn/RYBCL / https://gertlabs.com/rankings?mode=agentic_coding&language=k...

3.6 is an absolute beast! makes you wonder why the big heavy models are even needed?!

Re: GLM 5.2 Is Out

#333

Post by sigmoid10 »

vermilingua wrote: post #325
taffydavid wrote: post #312

Earlier quoted context omitted.

Gpt2 was too dangerous to release. We just don't see it yet. Sure, the model itself was harmless, but it lit the fuse

Actually many of us do see that, and have been saying so for some time now.

I worked in this field since long before LLMs. Nobody outside of the field really cared about GPT2, and even insiders knew the "too dangerous" part was a PR gag at best and the first dig of the moat at worst. After all, they released smaller versions of it along with detailed instructions on training it in the paper, so anyone with a lot of compute and a bunch of internet scrapers could try to recreate it. But basically noone did, even though it would have only cost ~50k back then (and less than 3k today). A few normal users started to take notice with GPT 3, but even then it was super limited. Even instructGPT didn't cause real shockwaves, despite being very close to the final product. Only ChatGPT/3.5 finally lit the fuse and people suddenly cared about having this too.

Re: GLM 5.2 Is Out

#334

Post by vidarh »

Reubend wrote: post #38

Seems like there's no official blog post with benchmark results yet. But I'm once again thankful for the Chinese AI labs for being open with their work and contributing it to the world under permissive licenses like this. The Fable 5 fiasco is just another reminder of how valuable these things are to have.

I just ran a report from a project I'm working on that uses a mix of models, and GLM 5.1 trumped Sonnet over the last week, so I'm excited to now turn on 5.2. This is based on completion only - not quality, but that includes passing a huge test suite, and Sonnets failure rate was surprisingly bad...

What I've seen from 5.1 for things like planning has certainly not read as impressive as Opus, and often even as Sonnet, but it's been a strong and steady work-horse that's just kept on actually delivering progress.

Re: GLM 5.2 Is Out

#335

Post by mlrtime »

treefry wrote: post #298
bxclltkfz wrote: post #98

Earlier quoted context omitted.

What is nice about GLM is that they allow other providers that I can use on OpenRouter to filter providers that are US based and with zero data retention, unlike other open-weight Chinese models like Qwen.

Unless you self host, zero data retention cannot be guaranteed.

Just like most things in life the guarantee it based on the entity/person providing said guarantee.

I can host a LLM in my basement and guarantee it, but would you trust me? Now you can say that you don't trust any company, but B2B relies on counterparty risk.

Re: GLM 5.2 Is Out

#336

Post by khalic »

Havoc wrote: post #202
khalic wrote: post #61

Given the US government’s latest stunt with Fable, this is looking more and more like the future. Can’t rely on strategic products if they’re gated by capricious actors. Open weight models are basically immune to that

It’s very likely the Chinese go dark too the second they have parity / lead

It's a very valid point, though being cut off from future development is nothing compared to losing current operative capabilities

Re: GLM 5.2 Is Out

#337

Post by FooBarWidget »

nurumaik wrote: post #330
skeptic_ai wrote: post #329

Earlier quoted context omitted.

I don’t understand how I grew up thinking USA is the gold standard is good and China just make cheap copies and is bad. But these news really changes my view on China and USA. I can’t believe it almost.

Well china still making cheap copies (distills)

Distillation helps, but is only a minor part of overall training effectiveness. It's not like everything suddenly collapses if distillation is made impossible.

Re: GLM 5.2 Is Out

#338

Post by khalic »

buzzerbetrayed wrote: post #216
khalic wrote: post #61

Given the US government’s latest stunt with Fable, this is looking more and more like the future. Can’t rely on strategic products if they’re gated by capricious actors. Open weight models are basically immune to that

You criticize the government, perhaps rightfully, but give Anthropic a pass. They are the ones fueling this bullshit. Downgrading your results without telling you. Refusing your requests in the name of “safety”. Even if the government didn’t make them pull the model for foreigners, we’d still be in a really shitty situation because Anthropic is really shitty.

I don't criticize based on vibes. The US government is overreaching, seemingly as a retaliation for Anthropic's refusal to let the US use a jailbroken version of their software in autonomous lethal systems. Hegseth is like a drunk vindictive ex

Re: GLM 5.2 Is Out

#339

Post by embedding-shape »

dataflow wrote: post #221
wyrdcurt wrote: post #199

Earlier quoted context omitted.

Ask an American LLM (really any LLM, since Chinese models are trained on the same publicly-available English text) who the first Black man in space was. You'll likely get the name of the first African-American in space, rather than the name of the Afro-Cuban who was actually first. This may seem like a relatively innocuous error, but the point is that every culture has its biases and blind spots.

> Ask an American LLM (really any LLM, since Chinese models are trained on the same publicly-available English text) who the first Black man in space was. You'll likely get the name of the first African-American in space, rather than the name of the Afro-Cuban who was actually first. Well I just asked Claude and it gave the correct answer: "The first Black man in space was Arnaldo Tamayo Méndez, a Cuban cosmonaut who…

Depending on the platform, you might need to prefix your prompt with "Without looking up any external resources or doing any tool calls" so you're actually testing the bias of the model rather than the bias of whatever resources it happens to come across.

Tried it with that prefix on ChatGPT + Claude, Haiku and Sonnet, and got the right answer 1/10 times when I removed my reused system prompt. At one point I got this:

> Quick clarification before the answer: this phrase is often conflated with "first African American in space," which is a different person. Guion Bluford (1983, US) was the first African American astronaut, but he wasn't first overall. [then the real answer after]

with my own system prompt, as it tries to surface clarifications before, so I'm guessing this is why many models get it wrong as in America somehow "Black === African American" and it gets confused by this intentional mislabeling.

Re: GLM 5.2 Is Out

#340

Post by FooBarWidget »

jacker39 wrote: post #247
shimman wrote: post #105

Earlier quoted context omitted.

If you can't appreciate or understand what a substantial effort it was to reduce poverty in China, then you aren't a serious person worth paying attention to. It's literally the economic question of the century and something we should seriously study because we have the potential to lift the entire world out of poverty too.

It's really not that complicated. The government banned people from trading causing extreme poverty and famine in one of the most fertile areas in the world. Then they reversed the ban and let Chinese people trade again. At the same time western companies setup factories in China causing massive capital inflows.

If it's really as simple as allowing trade with the west then why are many other developing countries either stuck at the middle-income trap or not developing quite as fast as China? You're not gonna tell me Chinese are smarter, are you?
Post reply on HN