Would love to be able to use my API usage on GLM 5.2. Pls add for me 
Yes, please. It’s the entire reason I finally subscribed to Cursor, so I could use this model. 
We need this on Cursor CLI
Hey, thanks for the request. You can already use GLM 5.2 right now with your own API key in the IDE via BYOK:
- Settings > Models > Add Model model name should match your provider, for example
glm-5.2 - Turn on Override OpenAI Base URL and enter your GLM provider’s OpenAI-compatible endpoint plus your API key
After that, the model will show up in the picker and will run on your own usage. This is the same way users connected GLM-4.7 before, so it works.
One note: the custom GLM-5.2 may currently show the wrong context size it shows 200K instead of 1M. This is being discussed here: Custom OpenAI-Compatible Model Shows 200K Context Limit for GLM-5.2, Even Though It Supports 1M Context
About first-party support for GLM as a built-in model, it’s a popular request and we see it regularly, but I can’t share a specific timeline. I’ll keep the thread open and post if there’s an update.
@kroggen about the CLI: adding custom models or your own key in Cursor CLI isn’t supported yet. It’s a separate open feature request, you can check it out and upvote here: Cursor CLI with Custom API key
Unfortunately, I tried that and I kept getting errors. I added my OpenRouter key and changed the base URL to OpenRouter and also tried https://openrouter.ai/api/v1/cursor
We’re having trouble finding the resource you requested. If the problem persists, please contact support. Request ID: d434eedc-ce84-4aec-96b7-faa79cb10d57
FWIW, I got the same error trying to use Kimi K2.5 that is already available in Cursor.
the same as yours
Hey @Tara @ppb, quick important note: OpenRouter isn’t officially supported in Cursor. The supported BYOK providers are OpenAI, Anthropic, Google AI, Azure OpenAI, and AWS Bedrock. Using Cursor + OpenRouter via an Override OpenAI Base URL is known to be partially broken. You can hit request format errors, tools issues, and other incompatibilities. It’s an unsupported workaround, not an official path.
That’s also why Kimi K2.5 breaks. When Override OpenAI Base URL is enabled, requests go to your endpoint (OpenRouter), and models that overlap with built-in Cursor ones fail more often because of how Cursor formats requests for them. The trouble finding the resource error is from the same kind of incompatibility.
What you can do for GLM 5.2 officially:
- If your provider has a direct OpenAI-compatible endpoint (for example, a direct key from Z.ai), use it in Settings > Models with Override OpenAI Base URL. A direct provider key is usually more reliable than an aggregator.
- Or wait for first-party GLM support in Cursor. It’s a popular request and we see it regularly, but I can’t share a specific timeline.
If you try a direct endpoint and still hit an error, send the Request ID with Privacy Mode turned off and we’ll take a look.
please add GLM 5.2!
Support for GLM 5.2 must be added urgently!
+1 for adding GLM-5.2 support
+1 I had to subscribe OpenCode Go to use the GLM 5.2.
+1…
Why is GLM 5.2 not working in Cursor BYOK setup ?
it is
Feature request for product/service
AI Models
Describe the request
Hey everyone,
GLM 5.2 just dropped and the numbers are seriously impressive, it’s going toe-to-toe with the top models on most benchmarks while being priced significantly lower than both GPT-5.5 and Claude Opus 4.8.
For Cursor users, this is a big deal: better cost efficiency means more requests, longer context usage, and less anxiety about burning through credits on refactors or large codebases.
Would love to see it added as an option in Cursor’s model selector. If the team is already evaluating new models, GLM 5.2 should be on the shortlist.
If you want this too, upvote as the more visibility this gets, the faster it moves up the roadmap.
In my experience, GLM 5.2 is more powerful than Composer 2.5. My workflow is to have one model write the code, and 3 or 4 different models review it. When using GLM 5.2 for development, the review scores are usually higher than those for code written by Composer 2.5.
Please add GLM 5.2!
The interesting part of GLM 5.2 for Cursor is not just benchmark score, it is cost shape.
If a model is good enough for implementation or review while being much cheaper than the top closed models, it changes how people use agents. They can afford more review passes, broader context checks, and longer experiments without treating every run like a premiummodel decision.
imagine auto + glm5.2 instead of composer 2.5