Try GLM-5.2: A Powerful Open-Weights AI You Can Access Right Now

A massive, freely available AI model just topped the charts - here's what that means for you.

What it is

GLM-5.2 is a text-based AI model released by Chinese AI lab Z.ai in June 2026. It was released under an MIT license, making the weights openly available. Independent benchmarkers at Artificial Analysis ranked it the top open-weights model on their Intelligence Index, ahead of other well-known open models. It is a 753B parameter, 1.51TB model - meaning running it locally requires substantial hardware that most users won't have access to.

Why it matters

Most of the best AI models are locked behind expensive subscriptions or closed systems - but GLM-5.2 is open weights, meaning the model weights are publicly available. It also ranked 2nd on the Code Arena WebDev leaderboard, behind only Claude Fable 5. Via OpenRouter, it is available from 9 providers at $1.40/million input tokens and $4.40/million output tokens, compared to GPT-5.5 at $5/$30 and Claude Opus 4.5-4.8 at $5/$25. If you want a capable AI for writing, reasoning, or coding tasks, this is worth knowing about.

How to use it

  1. 1Go to OpenRouter (openrouter.ai), which hosts GLM-5.2 through multiple providers.
  2. 2Search for GLM-5.2 in the model list - the source notes it is available from around 9 different providers on the platform.
  3. 3Type a text-based request - note that GLM-5.2 is text input only, so it cannot analyze images or photos you upload.
  4. 4Send your prompt and read the response. Be aware the model tends to produce longer, more detailed replies than other leading open models - Artificial Analysis found it uses significantly more output tokens per task.
  5. 5If you want to test its coding ability, try asking it to write or explain a small piece of code - the source highlights it ranks highly on front-end web development tasks.

Example

Simon Willison tested GLM-5.2 by asking it to 'Generate an SVG of a pelican riding a bicycle.' The model returned a fully self-contained, animated SVG illustration with working animations - the eyes and wheels all stayed in place correctly. He called it 'very impressive.' However, when he tried the same NORTH VIRGINIA OPOSSUM ON AN E-SCOOTER prompt he had used with GLM-5.1, the result was a step down - GLM-5.2 didn't even animate it. This mixed result is worth keeping in mind: the model impresses on some creative tasks but not all.

Practice it

Go to OpenRouter (openrouter.ai), select GLM-5.2, and ask it to 'Generate an SVG of a cat riding a skateboard.' Copy the SVG code it returns, paste it into a browser or an online SVG viewer, and check whether the animations work and the parts stay connected correctly. Then try a second animal or scene of your choice. This hands-on test gives you a direct feel for how GLM-5.2 handles creative, technical text generation - and lets you see for yourself where it impresses and where it falls short.

Source