For couple of days now, the custom GLM-4.7 with the Lite coding plan suddenly throws me: **We’re having trouble connecting to the model provider. This might be temporary - please try again in a moment.
**
Followed all the instructions on their website for connecting, was working previously, but it just randomly stops. No way near using my plan quota.
Request ID: c44efe79-4bdf-4ba2-b32a-9a77e28b8f5f
Hey, thanks for the report. Just to note up front, GLM-4.7 via the z.ai endpoint isn’t an officially supported setup in Cursor. The Override OpenAI Base URL option is meant for OpenAI-compatible providers, and we can’t guarantee it’ll work with custom endpoints like z.ai.
That said, there’s a workaround that has helped some users. Try using the model name in lowercase: glm-4.7 instead of GLM-4.7. More details here: Custom Model problems - #16.
Related threads:
Let me know if the workaround helps.
Hello, I am the one who suggested the lowercase model name. It is working indeed, but apparently glm-4.7 is the “behind the scenes” model by cursor itself, so it will be using your cursor credits.
More info here: https://forum.cursor.com/t/using-your-own-gemini-api-key-in-cursor-does-it-count-toward-cursor-limits/151397/11
Hey @ToX, great catch. Thanks for calling it out.
You’re right. Using the lowercase glm-4.7 makes Cursor apply its built-in model routing (via Fireworks) instead of your custom z.ai endpoint. So it works, but it isn’t actually using your BYOK setup.
And yes, Colin is actively digging into this in the related thread. The logs confirmed that GLM-4.7 BYOK requests are routed through Cursor’s API key, not the user’s.
I used GLM-5 to duplicate my android app into flutter. It analyze the code of the android app and make plan but upon execution it gives “Provider returned error: {“error”:{“code”:“1210”,“message”:“Invalid API parameter, please check the documentation.Request 143673 input tokens exceeds the model’s maximum context length 202752”}}”.
* I made subscriptopn to GLM coding Pro plan.
* Iam using https://api.z.ai/api/coding/paas/v4.
Is there a solution? How to do it?
Thanks.
I have the same issue. Everything worked perfectly until now. I think we should just wait.
This topic was automatically closed 22 days after the last reply. New replies are no longer allowed.
