Building a Linux Desktop from Scratch Using a self-review loop, GLM-5.1 spent 8 hours autonomously refining features, styling, and interactions to build a functional desktop environment.

SOTA on SWE-Bench Pro (58.4): GLM-5.1 delivers significant leaps in coding and agentic performance.

Introducing GLM-5.1: The Next Level of Open Source - Top-Tier Performance: #1 in open source and #3 globally across SWE-Bench Pro, Terminal-Bench, and NL2Repo. - Built for Long-Horizon Tasks: Runs autonomously for 8 hours, refining strategies through thousands of iterations. Blog: Weights: API: Coding Plan: Coming to in the next few days.

Vector-DB-Bench: 6x Performance Boost In high-performance database optimization, GLM-5.1 reached 21.5k QPS over 600+ iterations and 6,000+ tool calls. This is 6x the performance of a standard 50-turn session.

Special thanks to our launch partners, AI gateways, and inference providers. Access GLM-5.1 now: - OpenRouter: - Vercel: - Requesty:

yeah everything's fine till this one. Main one as orchestrator + 3 subagents and they just face with requests limit on the max plan

8 hours of self-review to build a linux desktop is kind of insane. anime training arc but for window managers

What method or harness are you using for the self-review loop?

which self review loop? Could you specify this?

How much $$$ in tokens for that? 👀

What harness was this using? Like cloud code or open code or droid or something like that?

can somebody tell me what's the app where is glm 5.1 used in this video? Glm has a dedicated desktop app? or ? ty

Wtf this is a new distro or what?

Noe can you make auto claw work with linux?

@Zendiyy really impressive 🔥🔥

Wow, this is the Most goted video of a model release

#fail it built macOS 😅 jk

Is it really ? I don't believe it

This is more than awesome.

this is cool

@theo 0:32 🙌🙌

wt @grok is this real ?

what is a self-review loop? can you share more about this ?

Wow.

8 hours of autonomous iteration to build a full desktop is actually wild 👀 feels like we’re getting closer to AI handling real end-to-end systems, not just snippets or demos. Curious how polished it actually is in daily use 🔥

wow

Let's see the code