alt.hn

6/16/2026 at 9:07:04 PM

GLM-5.2: Frontier Intelligence, Open Weights

 https://twitter.com/Zai_org/status/2066938937344495629

by zixuanlimit

6/16/2026 at 10:12:35 PM

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.

by osti

6/16/2026 at 10:19:46 PM

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.

by LaurensBER

6/16/2026 at 11:10:40 PM

> it does do design extremely well

Better than Opus?

by ebbi

6/17/2026 at 2:50:56 AM

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

by osti

6/17/2026 at 6:39:46 AM

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.

by ebbi

6/17/2026 at 12:46:47 AM

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.

by fcanesin

6/17/2026 at 12:24:30 PM

IPO within year of founding?

It it normal for startup in China?

by saxenaabhi

6/16/2026 at 9:59:42 PM

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...

by dbbk

6/17/2026 at 12:20:44 AM

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.

by entrope

6/17/2026 at 12:21:52 AM

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.

by api

6/17/2026 at 3:31:44 AM

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

by _aavaa_

6/17/2026 at 8:25:42 AM

Aren't Macbooks limited to 128GB RAM?

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

by Tepix

6/17/2026 at 12:12:08 PM

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

by _aavaa_

6/17/2026 at 2:33:34 AM

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

by coolspot

6/16/2026 at 9:36:54 PM

I have already tested this and it is impressive.

by Bucko1