Cannot use custom models for inline edits

[Pear2565]

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I cannot change the model used for inline edits. On my work laptop, it’s stuck on deepseek. On my work desktop, it’s stuck on composer. Adding new models via the drop down does not fix it.

Issue is present on MacOS and Linux machines (tried on 2 separate systems).

Steps to Reproduce

1. Press ctrl + k
2. Press add models
3. Add a model
4. The model that you added does not appear on inline edit

Expected Behavior

Adding a new model via the ‘add a model’ menu should add the model to the inline edit panel, so that users can use the selected model for editing.

Screenshots / Screen Recordings

Operating System

MacOS
Linux

Current Cursor Version (Menu → About Cursor → Copy)

Version: 2.1.19
VSCode Version: 1.105.1
Commit: 39a966b4048ef6b8024b27d4812a50d88de29cc0
Date: 2025-11-21T22:59:02.376Z
Electron: 37.7.0
Chromium: 138.0.7204.251
Node.js: 22.20.0
V8: 13.8.258.32-electron.0
OS: Darwin arm64 24.5.0

Additional Information

This is blocking me for using inline edit completely. I don’t want to use deepseek (too incompetent) or composer (too competent but too expensive).

For inline edits, I’m doing small things: adding a subplot to a matplotlib plot or doing some kind of calculation. I need a model that is smart and cheap. Namely: gemini 2.5 flash or grok code fast 01, which are both free and good for this type of thing.

Does this stop you from using Cursor

Yes

[deanrie]

Hey, thanks for the report. This sounds frustrating, especially since it’s blocking your workflow.

FYI, in Cmd+K inline editing only “non-thinking” models are shown, and you can enable the others in Cursor Settings.

[rollo]

Most models now are labelled “thinking” in Cursor even the hybrid ones and some are suited for quick inline edits/questions, very specific example: Gemini 2.5 flash which cannot be used in Cursor, unless we have to enable/disable Google API for BYOK every time.

@deanrie what does the Cursor team think about it?

Gemini 2.5 flash is the default inline model in Google’s Antigravity IDE by the way.

[Pear2565]

Hi @deanrie, thanks for your reply. How do I enable the thinking models to show in inline edit? This is all that I see in the settings for inline edit:

I believe thinking models should be included in the inline edit menu by default. It’s not clear to me why it would be different.

The only other non thinking models that I see are: GPT 4.1 and Kimi K2 Instruct. Are either of those 0requests like Gemini Flash/Grok Fast?

[deanrie]

Thanks for the question. I see you want to use Gemini 2.5 Flash or Grok Fast with Cmd+K, but they don’t show up.

The issue is that Cmd+K (inline edit) only lists non-thinking models by default. In Cursor, Gemini 2.5 Flash and Grok Fast are marked as thinking models, so they’re hidden.

There’s no setting to enable thinking models in Cmd+K right now - that’s a design limitation. Available non-thinking models include: GPT-4.1, Kimi K2 Instruct, DeepSeek, Claude Sonnet.

Workarounds:

• Use Chat Cmd+L with Gemini Flash or Grok Fast, then apply code manually
• Or try GPT-4.1 or Kimi K2 for inline edit

I’ll pass this feedback to the team - many users are asking to bring back thinking models in Cmd+K or make it a setting.

[rollo]

Yes, thanks @deanrie for forwarding. I mean, it won’t bother users.

You could make it an experimental setting, “enable thinking models for inline chat”, at least we can get Gemini flash 2.5, which is pretty good at that.

So that everyone is satisfied and have a choice to enable or not

[Pear2565]

Thank you @deanrie please let us know what they think!