GLM-5.2: Frontier Intelligence, Open Weights

Posted by zixuanlimit 15 hours ago

Counter32Comment19OpenOriginal

Comments

Comment by osti 14 hours ago

Given that DeepSwe is one of the very few coding benchmarks worth taking a look at, this achieves rather excellent result at it (not far from opus 4.8).

From looking at the results and my own impression of 5.1 and other models, I think this is the best Chinese coding model by some non-insignificant margin.

Comment by LaurensBER 14 hours ago

I've been very pleased with it's performance over the last few days.

It's definitely not near Opus 4.8 level but it's very impressive nonetheless and it does do design extremely well.

Comment by ebbi 13 hours ago

> it does do design extremely well

Better than Opus?

Comment by osti 10 hours ago

I don't know what people mean when they say design lol, is it for frontends?

Comment by ebbi 6 hours ago

Yeah, that's what I mean anyway. Each model has certain design tropes it repeats everywhere, and some of them are very old-school or not really UI best practice.

And then the more ambitious cases where you ask for a feature without being prescriptive with UI needs, the end result is sometimes atrocious with weird font use, colours, etc.

Comment by fcanesin 12 hours ago

Zhipu AI is founded by a superstar Tsinghua professor, did an IPO in January (Hong Kong stock exchange) hired half it's past research lab and it's stock is >10x since. This is not a "just distill Claude" thing.

Comment by saxenaabhi 38 minutes ago

IPO within year of founding?

It it normal for startup in China?

Comment by hadik 14 hours ago

Blog post with more information including benchmarks: https://z.ai/blog/glm-5.2

Comment by gnabgib 11 hours ago

Discussion (765 points, 3 days ago, 489 comments) https://news.ycombinator.com/item?id=48518684

Comment by coolspot 10 hours ago

Looks like their code subscription is sold out. Clicking on the "Subscribe" button just does nothing. Pay-as-you-go API works well.

Comment by dbbk 15 hours ago

If I have a fully maxed out MacBook Pro, would it make sense to just switch from Opus 4.8 to this? I've never tried running local models for coding...

Comment by entrope 12 hours ago

HuggingFace says this model has 753B parameters, which will need a lot more RAM than a maxed-out MacBook Pro. With 40B active parameters, running from SSD would need patience.

Comment by _aavaa_ 9 hours ago

For an fp4 quantization it should fit with room to spare for KVCache

Comment by Tepix 4 hours ago

Aren't Macbooks limited to 128GB RAM?

FP4 would require >350GB RAM + KV cache, so no.

Comment by _aavaa_ 51 minutes ago

Oops, you’re right. My brain understood Mac Studio for some reason.

Comment by api 12 hours ago

I’ve wondered for a while if anyone is working on very wide channel parallel (kind of like RAID 0) SSD for this purpose. Couple that with a tensor processor and that would be interesting.

Comment by petu 4 hours ago

There's talks about HBF. F for Flash -- HBM packaging and bus width, but using NAND memory.

e.g. https://www.sandisk.com/company/newsroom/press-releases/2026...

Comment by Bucko1 15 hours ago

I have already tested this and it is impressive.

Comment by ChrisArchitect 11 hours ago

non-x blog post: https://z.ai/blog/glm-5.2

And previously: https://news.ycombinator.com/item?id=48518684