01What is GLM 5.2 vs Claude?

GLM 5.2 vs Claude is a direct comparison of two AI models from China and the US. GLM 5.2, made by Zed (Zhipu AI), competes with Claude (Opus 4.8) from Anthropic. In this episode, Julian Goldie tested both models side by side with the same prompts.

He gave them five different tasks: a Temple Run game, a solar system simulation, a liquid particle effect, a landing page, and a Neon Arcade game. The results show which model creates better, more fun, and more polished output. GLM 5.2 won most tests, but not all.

02Main Features of GLM 5.2

GLM 5.2 is a brand new AI model that just dropped. It has a million token context window, which is huge. That means it can handle a lot of information at once.

  • It's from China, made by Zed (Zhipu AI).
  • It can be plugged into AI agents like Hermes.
  • It works with Claude desktop for orchestration.
  • It's cheaper on subscription than Claude.
  • It supports reasoning, similar to other top models.

Because it's so new, it wasn't on OpenRouter at the time of testing. Only GLM 5.1 was available there. This means official benchmarks are not yet published.

03How to Set Up GLM 5.2

To use GLM 5.2, you need to integrate it into your AI workflow. Julian showed how he sets it up in his agent operating system.

  1. Create an account or subscription for GLM 5.2 (likely through Zed).
  2. Access GLM 5.2 via API or the provider's platform.
  3. Connect it to your agent tool like Hermes or Claude desktop.
  4. Configure the model settings, like context window and reasoning.
  5. Test it with simple prompts before full use.

Julian uses Claude desktop as the main controller, then plugs GLM 5.2 into the agent operating system. This lets you orchestrate multiple models together.

04How to Use GLM 5.2 for Building

You can use GLM 5.2 to create games, simulations, and landing pages. Julian showed several examples.

  • Temple Run game: GLM made a fun, intense runner with dodging and speed ramps.
  • Solar system map: It created an animated orbit map, but not the best one.
  • Liquid in a bowl: It made a interactive particle effect with mouse tilt and theme changes.
  • Apple-style landing page: It built a nice page with working menu and scroll reveals.
  • Neon Arcade game: It produced a crazy, fun game with sound effects.

To get these results, you give a clear, detailed prompt. For example, the liquid test prompt was: "thousands of particles sloshing round in a bowl. You tilt with the mouse.

Soft glowing metaball look."

05Results and Comparisons from the Tests

Julian ran five tests and compared GLM 5.2, Kimi K 2.7, and Claude (Opus 4.8). Here are the winners:

  • Temple Run: GLM 5.2 won – it was the most fun.
  • Solar system: Kimi K 2.7 won – it had zoom, pan, and speed controls.
  • Liquid in a bowl: GLM 5.2 won – it was the most interactive.
  • Landing page: GLM 5.2 won – it had a working menu and nice UI.
  • Neon Arcade: GLM 5.2 won – it was way more fun.

So GLM 5.2 won 4 out of 5 tests. Kimi K 2.7 won one. Claude didn't win any.

Julian was surprised that GLM 5.2 beat Claude so often.

06Tips, Limitations, and Mistakes to Avoid

No AI model is perfect for every task. In the solar system test, GLM 5.2 didn't allow zooming, which was a downside. Claude's landing page had no menu reveal, and its arcade game felt buggy.

  • Don't assume one model wins everything – test for your specific need.
  • Check if the model is available on your platform – GLM 5.2 wasn't on OpenRouter yet.
  • Consider the subscription cost – GLM 5.2 is cheaper than Claude.
  • For agents, you can use GLM 5.2 and Kimi, but not Claude – that's a big difference.
  • Use Claude desktop to orchestrate all models together for best results.

Julian's personal method: use Claude desktop as the main operator, then plug GLM and Kimi into the agent system. That way you get the best of each.

07Who Should Use GLM 5.2?

GLM 5.2 is for anyone who builds AI-powered projects, like games, websites, or simulations. It's especially good if you want high-quality output at a lower cost than Claude.

If you use AI agents, GLM 5.2 is a key model because you can integrate it into tools like Hermes. That opens up more flexibility than Claude, which can't be used in agents.

Julian recommends trying it for creative builds. In the tests, it consistently made the most fun and polished results. Whether you're a developer, hobbyist, or business owner, GLM 5.2 is worth checking out.

08Where to Get the Agent Operating System

Julian built a complete agent operating system that includes GLM 5.2, Kimi K 2.7, and Claude. It has features like Obsidian memory, a memory galaxy, and daily updates.

You can get it from the AI Profit Boarding at AIProfitBoarding.com. The link is in the video description and comments.

Inside the community, you get video tutorials, a changelog, zip install files, and access to Julian who personally answers questions. There are also four weekly coaching calls where you can share your screen and build with others. You can even find people in your local area who use AI agents.