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

GLM 5.2 beats Claude in our benchmarks

semgrep.dev

Post reply on HN
Print view View original

551–559 of 559 posts

Re: GLM 5.2 beats Claude in our benchmarks

#551

Post by unnouinceput »

throwaway676712 wrote: post #550
unnouinceput wrote: post #540

Earlier quoted context omitted.

I was 100% sure that somebody will throw this, but I didn't actually expect to do it from a throwaway account. Maybe because you know calling Android Linux is like calling a human just an ape (but mirrored because you know, Android in this case is the ape). Oh, and Microsoft Loves Linux, right? Because that's why they invented WSL, to make people go and use Linux, right? riight!!

Moving goalposts. Android is Linux (the GNU part of GNU/Linux is trivial to add), people use Linux knowingly or not, and Linux is on the most devices on Earth regardless of whether their users know it or not.

By that logic, billions of users are using OS/2. Or...wait for it...everybody is using Unix.

Re: GLM 5.2 beats Claude in our benchmarks

#552

Post by naasking »

throwawayffffas wrote: post #509
naasking wrote: post #501

Earlier quoted context omitted.

Why? 4 x DGX sparks should be enough. That's way less than $80k.

From a quick google search a DGX spark seems to decode Llama 3.1 70B (FP8) at 2 tokens per second. I would expect the performance on a 768B parameter model spread across 4 to be significantly lower even though its a mixture of experts. For real work anything below 60 tokens per second is essentially unusable. That's not taking into account the prompt filling, Llama 3.1. 70b on DGX spark runs at about 800 tps running…

> I would expect the performance on a 768B parameter model spread across 4 to be significantly lower even though its a mixture of experts.

To fit that model you'd need multiple machines which lets you run tensor parallelism for additional speedups, roughly X speedup for X machines. Also you can't naively compare tps across models, and it would certainly run a nvfp4 quant or similar.

Re: GLM 5.2 beats Claude in our benchmarks

#553

Post by naasking »

villish wrote: post #533
naasking wrote: post #501

Earlier quoted context omitted.

Why? 4 x DGX sparks should be enough. That's way less than $80k.

It would run so slow it would be functionally unusable or at such a low quant that it wouldn't be useful for serious work.

nvfp4 is pretty good, and with 4 machines you get a roughly 4x speedup using tensor parallelism.

Re: GLM 5.2 beats Claude in our benchmarks

#554

Post by frr149 »

pimeys wrote: post #59

I have taken another look on these open models after the fiasco of Fable and GPT 5.6 this weekend and... GLM-5.2 truly is a good workhorse model for daily programming. I consider myself a heavy user of LLMs and a seasoned developer. A typical session for me with GPT is usually over a hundred dollars... This weekend I programmed a matrix bot with encryption and a Rust agent with some tools. Because I need one and Open…

How are you using it? A subscription from z.ai?

Re: GLM 5.2 beats Claude in our benchmarks

#555

Post by throwaway676712 »

unnouinceput wrote: post #551
throwaway676712 wrote: post #550

Earlier quoted context omitted.

Moving goalposts. Android is Linux (the GNU part of GNU/Linux is trivial to add), people use Linux knowingly or not, and Linux is on the most devices on Earth regardless of whether their users know it or not.

By that logic, billions of users are using OS/2. Or...wait for it...everybody is using Unix.

No

Re: GLM 5.2 beats Claude in our benchmarks

#556

Post by nozzlegear »

fragmede wrote: post #130
techpression wrote: post #56

Earlier quoted context omitted.

Reaper and Predator are both drones and there’s really no comparison to toy drones in terms of sheer destruction and capabilities in general, the comparison is actually quite apt imo.

Which ones are the ones Ukraine has used to bomb Moscow?

Ukraine would absolutely use Reaper and Predator drones if they had access to them.

Re: GLM 5.2 beats Claude in our benchmarks

#559

Post by andai »

gertlabs wrote: post #172
pimeys wrote: post #59

I have taken another look on these open models after the fiasco of Fable and GPT 5.6 this weekend and... GLM-5.2 truly is a good workhorse model for daily programming. I consider myself a heavy user of LLMs and a seasoned developer. A typical session for me with GPT is usually over a hundred dollars... This weekend I programmed a matrix bot with encryption and a Rust agent with some tools. Because I need one and Open…

GLM 5.2 is a great model, but if you only want to use the best model available, it isn't there yet. Every lab releases models that memorize benchmark answers, both intentionally and unintentionally. But we consistently find that models from Chinese labs have a wider gap between public benchmarks and our evaluations, which we designed to be less vulnerable to benchmaxxing. In multi-agent coding environments, GLM 5.2 i…

> But when factoring in performance/cost, GLM 5.2 is the frontier model.

Isn't DeepSeek half as good and 20x cheaper?

"Half as good" sounds bad but for most trivial tasks I've found it more than good enough. (Why would I want to waste my Frontier AI Tokens on those?)

Post reply on HN