Subagents can only be run using expensive models (composer-2.5 vs composer-2.5-fast)

Support Bug Reports
, , ,
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

I have several flows with orchestrator agents / skills that spawn subagents doing smaller parts of the flow. Some of those i’d like to do as cheap as possible, so i pinned the subagent model to composer-2.5. I noticed it still started up as composer-2.5-fast.

After some debugging (the attached screenshot) using hooks to look at subagent_model at startup, it appears it’s not possible to start a subagent with composer-2.5. It will either fail to start or default to composer-2.5-fast.

Steps to Reproduce

Explicitly tell cursor to start a subagent using composer-2.5.

Expected Behavior

Subagents are able to start using composer-2.5.

Screenshots / Screen Recordings

image.png
image.png819×421 16.9 KB

Operating System

Windows 10/11

Version Information

Version: 3.7.42 (user setup)
VS Code Extension API: 1.105.1
Commit: 5702c9cfca656d8710fad58402fe37f14345e3a0
Date: 2026-06-15T19:39:42.738Z
Layout: editor
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
xterm.js: 6.1.0-beta.256
OS: Windows_NT x64 10.0.26200

Does this stop you from using Cursor

No - Cursor works, but with this issue

4

image
image846×614 40.4 KB

6

Hi @vorantz!

If you’re defining your own subagents, you should be able to use

model: composer-2.5[fast=false]

or

model: composer-2.5[]

in the frontmatter to resolve the non-fast variant.

We have a bug open on our side about this.

9

Hi, thanks for the the fast response.

Yes, i forgot to specify these are all custom subagents and i’m trying to set their model explicitly.
I tried the composer-2.5[fast=false] syntax and yes, I’m happy to say it works now. Thank you.

Is there any documentation where I could find this info for the future (the syntax and the possible model ids)? I can only find docs that point to other docs and lead to a dead end. Cursor itself also wasn’t able to find this info it seems.

11

Ideally, we will fix the bug so that composer-2.5 resolves to Composer 2.5, and composer-2.5-fast resolves to Composer 2.5 Fast!

That said, it’s weird the bracket syntax isn’t documented anywhere (and it’s useful for other models as well, like picking effort level).

claude-opus-4-8[effort=high]

Will flag to the team to see if we’re ready to document this!

14

Hey @vorantz @jojolionss!

FYI, specifying composer-2.5 as a slug for a custom subagent should now resolve to the “base” variant, not fast!