01What is GLM 5.2 and how does it work?

GLM 5.2 free is an open-source AI model you can use without paying, thanks to ZAI. It's a frontier-level model that often beats top competitors like Opus 4.8. You can test it, build with it, or even self-host it.

GLM 5.2 has a million-token context window, meaning it can handle huge amounts of text at once. It was released on June 14th with the tagline "the never forgets agent." The model is open weights, so you can download it from Hugging Face or run it free on ZAI.

In this episode, Julian Goldie showed how to access it. He also demonstrated its power by building games and an operating system with it. The key is that it's both free and powerful, which is rare.

02Main features of GLM 5.2

GLM 5.2 comes with several standout features that make it a top choice for many tasks.

  • Million-token context window: It can process very long documents or conversations in one go.
  • Open source: You can download the weights and run it locally if you have the hardware.
  • Free access on ZAI: No cost to try, though token limits apply.
  • Deep think mode: You can switch between fast answers and deeper reasoning.
  • Web search and image attachment: Enhance your queries with real-time info or images.
  • Agent support: Use it to build tools and apps in the background.

Julian highlighted that it outperforms GPT-5.5 and Gemini 3.2 Pro on his tests. It also beat Fable 5 on Design Arena, which is a big deal.

03How to get started with GLM 5.2 for free

Getting started with GLM 5.2 free is simple. Follow these steps to start using it today.

  1. Go to ZAI (set.ai) and create an account if needed.
  2. Select the GLM 5.2 model from the available options.
  3. Choose between the agent or chat mode, depending on your task.
  4. Optionally, enable web search or attach images to your prompt.
  5. Adjust settings like high or max token usage, and toggle deep think if needed.
  6. Start typing your prompt and hit enter to see results.

Julian showed this process live. He tested it with a simple question and then built an SEO tool using the agent mode. The agent worked in the background, taking about 10-15 minutes.

04How to use GLM 5.2 for various tasks

You can use GLM 5.2 for many creative and practical projects. Julian demonstrated several examples in the episode.

  • Building games: He created full 3D games like Neon Drift, a racing car game, and an open-world game with a single prompt.
  • Creating websites: He made a landing page and a voxel game runner, showing clean UI design.
  • Developing an operating system: GLM 5.2 built a full OS with apps like notes, terminal, and a music player. It even saved data between sessions.
  • Generating music: Using agent OS, he created music that played through the GLM-built OS.

For best results, Julian recommends using GLM 5.2 via CLI in an agent operating system. This makes it faster and more efficient than the ZAI sandbox, which can be slow.

05Results, tests, and comparisons from the episode

Julian ran multiple tests and comparisons showing GLM 5.2's performance. On his Goldie Benchmarks, it scored 8.23 out of 10 across 21 tasks. It won several medals and was released just last week.

In side-by-side tests with Opus 4.8, GLM 5.2 produced more fun and visually appealing outputs. For a website design, GLM 5.2's version felt cleaner and nicer. On Design Arena, it beat Opus 4.8 and even Fable 5, which was so powerful it was taken down.

Comparisons with Kimi K217, another Chinese open-source model, also favored GLM 5.2. The outputs were more engaging, with better animations and themes. Julian noted that GLM 5.2 has caught up to Opus 4.8 quickly, calling it a "DeepSeek moment for 2026."

06Tips, limits, and mistakes to avoid with GLM 5.2

While GLM 5.2 is impressive, there are some things to keep in mind. Julian shared his experience and advice.

  • Speed: The free ZAI version is very slow, especially with deep think. Use CLI or self-hosting for faster performance.
  • Token limits: Free access has token limits, so plan your usage.
  • Hardware requirements: Running it locally needs a powerful setup, though people have run quantized versions on a Mac Studio.
  • Chinese model concerns: Some users may have privacy or trust issues with a Chinese model. Consider that when choosing.
  • Best use case: GLM 5.2 excels at UI tasks, so use it for design, games, and creative projects.

To avoid mistakes, use it in an agent OS for better speed. Also, test it side by side with other models to see which works best for your needs.

07Who should use GLM 5.2 and why it matters

GLM 5.2 is for anyone who wants a powerful AI model without paying. Developers, designers, and creators can all benefit from its abilities. It's especially useful for those building apps, games, or websites.

Why it matters is that it shows open-source models can compete with closed ones. Julian pointed out that once open weights are out, they can't be taken down, unlike closed models like Fable 5. This changes the AI landscape.

In this episode, Julian also mentioned his community, the AI Profit Boardroom, where you can get full setups and tutorials. But even without that, you can start using GLM 5.2 free right now.