Requested subagent model is sometimes unavailable locally

Support Bug Reports
1

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Sometimes Cursor reports that a requested subagent model is unavailable locally, even though the model becomes available immediately after switching the active model to that requested model and retrying.

Steps to Reproduce

  1. In Cursor, ask the agent to create a PR and review it using a specific high-end model, e.g. gpt-5.5-high.
  2. Cursor creates the PR.
  3. Cursor says it cannot spin up a subagent on gpt-5.5-high because the only available subagent model is something else, e.g. composer-2.5-fast.
  4. Switch the active model in Cursor to gpt-5.5-high.
  5. Retry the same request.
  6. Cursor now recognizes gpt-5.5-high as available and starts the review subagent.

Expected Behavior

When I request an agent/subagent using a specific model, Cursor should either:

  1. Correctly detect that the requested model is available, or
  2. Refresh available models automatically before deciding it is unavailable.

The availability of a subagent model should not depend on manually switching the main active model first.

Operating System

MacOS

Version Information

Version: 3.8.11
VS Code Extension API: 1.105.1
Commit: e56ad3440df06d22ca7501e65fd518e905486ef0
Date: 2026-06-18T01:40:18.333Z
Layout: glass
Build Type: Stable
Release Track: Default
Electron: 40.10.3
Chromium: 144.0.7559.236
Node.js: 24.15.0
V8: 14.4.258.32-electron.0
xterm.js: 6.1.0-beta.256
OS: Darwin arm64 25.5.0

For AI issues: which model did you use?

GPT 5.5 High

For AI issues: add Request ID with privacy disabled

c4009c42-ec85-4c50-8e7b-05d5bc659926

Does this stop you from using Cursor

No - Cursor works, but with this issue

6

Hey @EricWiener,

This is known behavior rather than a backend failure: the models a subagent can be spawned with come from the models you have enabled and visible in your main model picker (Cursor Settings → Models). gpt-5.5-high wasn’t in that set, so it wasn’t offered to the subagent. Selecting it as your active model is what adds it to that enabled/visible list, which is why switching and retrying “works.”

To get it working without switching your active model:

  1. Open Cursor Settings → Models.
  2. Make sure GPT 5.5 is Enabled and visible in your model picker, and set to the reasoning level you want (e.g. high). Visibility is what feeds the subagent list, not just being enabled, and a subagent gets whatever reasoning level that family is currently set to in your picker.
  3. Start a new chat and request the subagent on that model again.

Once it’s enabled and visible, it’ll be offered to subagents directly, no active-model switch needed. For reference, the model field for custom subagents takes a family-level ID like gpt-5.5 and follows the same picker setting.

8

Hi @mohitjain this works fine 95% of the time I ask the specific gpt 5.5 high sub agents. I ask for the same review sub agent every time and I don’t change sub agents. I don’t think the 5% is functioning as intended.