GLM-5.3 cannot be implemented reasoning_effort

Where does the bug appear (feature/product)?

As the title suggests, I have added Z-AI GLM-5.3 to the model in the cursor using the Openai key, but I found that it cannot choose the reasoning effort and fast modes like other large models

Hi @ran_shao Thanks for the post. What you’re seeing is expected rather than something going wrong.

The reasoning effort and Fast controls are driven by Cursor’s model catalog, so they only appear for models we serve. When you add a model by typing its name and supplying your own OpenAI-compatible key, Cursor has no parameter definitions for it, so those selectors are hidden rather than shown with no effect behind them.

For a custom GLM-5.3 endpoint, the reasoning behavior has to be configured on the provider side. The usual approach is to rely on the endpoint’s own default, or to point Cursor at a model ID variant that already encodes the thinking mode you want.

Thanks

Hi i have a idea what if u allow setting thinking paramteres for third party models too?