I downloaded it. I tested it. I'm blown away.

Frontier AI is incredible.

But it's slow and expensive. Plus, everything you send it gets stored on someone else's servers.

GLM 5.2 just changed that.

A few weeks ago, Unsloth released a version of this model you can run locally on your own machine. Near Opus 4.8-level intelligence, completely free, completely private, running 24/7 on your desk.

I ran it on my Mac Studio and I am completely blown away.

In case you missed it, I put together a full breakdown of everything you need to know.

Here's what I cover in the video:

• ​What GLM 5.2 actually is and why it's comparable to Opus 4.8 (benchmarks are one thing, but the 3D first-person shooter it built for me told me everything I needed to know)
  ​
• ​Local vs. frontier: when to use each one (the simple passive vs. speed framework I use to decide which model gets the job)
  ​
• ​What hardware you actually need (and what to run if you don't have a Mac Studio)
  ​
• ​How to set it up in one step using Hermes (no technical knowledge required. I just sent my agent a link and it handled everything)

Jump to whatever section is most relevant. Or just watch the whole thing.

​Watch the full breakdown here.​

Let me know what you're running it on.

Keep shipping,

Alex