Truly unbelievable GLM 5.2 just released and it's an open weights model you can run locally The insane part is, it's just as good as Opus 4.8 Unlimited, free super intelligence running on your desk In this video I cover how it works, and how to set up your... first local model:show more

One problem: I didn't have the chance to listen to you back in January and now Mac studios are so dang expensive!

people called it all hype! now you can't get studios anywhere

I wish I was around to hear you say that because I would have got one but yeah now if I were to buy just one of the basic 96gb m3 ultra's that's quite a bit of money especially for me.

Yep, unbelievable! Just turn on your laptop with 400GB of vram, and watch it run GLM 5.2 4-bit.

While you technically *can* run it locally, you still need a beast desktop setup worth at least 10k

Running on your desk” is doing a lot of work here 😄 753B at usable speed needs a multi-GPU server or a maxed Mac Studio — for most devs “local” still means an API, just 70% cheaper. The open-weights part is the real win. The “free superintelligence on a laptop” part isn’t here yet.

I'd say it's a good as Opus 4.6.

free unlimited superintelligence on a desk is the kind of phrase that should make every closed lab nervous. timing right after the fable mess is perfect. the moat people thought existed is melting faster than anyone predicted

Hard pass on China LLM. Our gov just restricted our own company's LLM with Anthropic. You don't think our gov will ultimately restrict use of state sponsored Chinese LLM? Building on Chinese LLM in the US is setting yourself up for failure.

There's absolutely no way quantized version of GLM 5.2 is even close to Opus 4.8, and no way it's even remotely as fast.

Another easy bookmark. Thank you Alex. How is it compared to opus from a creative standpoint? Ie. Content writing.

it's up there creatively too. good frontend

Awesome, appreciate you taking the time Alex.

Why do people post this horseshit? To run GLM locally you'd need about $750k worth of computer.

Appreciate all your content

Calling it super intelligence makes you look like a clown.

I only have 1 dgx spark. So those of us that don’t have the compute you have can’t run local

glm 5.2 is scary good guys! with it 127-employee z ai beats 180k-employee google on agent arena 🤯

That day you said something back in January, I actually went all in on AI as well. I ordered a few mac mini's and a 512gb studio immediately, got into openclaw right away then hermes (Best thing ever) and now going local... Best decision I've ever made and a huge thanks to you brother.

makes me very happy to read brother!

I just started using GLM 5.2 in Codex through Ollama. It is in every bit as good as GTP 5.5. Very impressive!

we miss u bro <3 thanks for looking out even in grind mode

ill be back soon man. gotta get this beta launched

GLM-5.2 top 1 open source ai model now great new update from their team

I need a computer run local models. OpenAI putting me at 💯 usage all week out of the blue was the last straw. They have been cancelled.

GLM is just like backup for most of us

We are testing GLM 5.2 with @/Anchored’s local assistant. @/anchored, the most intelligent Notes app in the world!

I am currently trying to transfer it to a office laptop runnable model, super interesting stuff

@grok, realistically, what's the cheapest hardware to run GLM 5.2 at home?

Nice post but very little people have $10,000 around to get decent hardware to be able to run it...

I'll save you some time. A 18:31 long video, half the way he's still talking about "what is a local model".

open source just keeps getting stronger results

It's better than opus 4.8*

I'm waiting for the new laptops with NVIDIA RTX Spark so I can run GLM models

The leap from one benchmark to 'unlimited free super intelligence' is truly riveting.

running locally just means no more cloud latency on my rigs

Its a great model but you sure aren't going to run this on your desk unless its a massive workstation of a mac studio.

AI泡沫最危险的时候是所有人都觉得它能解决一切。现实是：它连很多简单问题都搞不定。 「2 刚刚发布，它是一个开源权重的模型，你可以在本地运行」

Which config it takes to run !

@AlexFinn did you cancel your Codex chat GTP 5.5 subscription? I got codex 5.5 because you recommend it. Please advice. Thank you! you are the best!

This is even stronger than the hype, GLM-5.2 just took **#1** on the HTML web design leaderboard, beating Claude Fable 5 and Opus versions. And the pricing is ridiculous: ~$1.40/M input vs $10 for Fable 5.