Ollama model Qwen 3.6:35b stuck on model selection in codex after --restore. Codex not working at all.

What version of the Codex App are you using (From “About Codex” dialog)?

Version 26.623.81905 • Released Jun 30, 2026

What subscription do you have?

Plus

What platform is your computer?

Windows 11 Pro

What issue are you seeing?

Codex App is stuck on Ollama model qwen3.6:35b with no supported OpenAI models available. Codex is unusable.

Steps already tried:

• Ran ollama launch codex-app --restore
• Removed Ollama/Qwen references from C:\Users\tatyf\.codex\config.toml
• Deleted/rebuilt C:\Users\tatyf\AppData\Local\OpenAI\Codex
• Uninstalled and reinstalled Codex App
• Signed out and back into ChatGPT Plus account
• Confirmed ollama ps shows no active model
• Qwen still appears as the default model and no GPT/OpenAI models are available

Current error:
“The 'qwen3.6:35b' model is not supported when using Codex with a ChatGPT account.”

What steps can reproduce the bug?

Feedback ID: 019f2054-ccb6-7342-a68e-6ac6c061d931

What is the expected behavior?

_No response_

Additional information

The following manual steps restored normal Codex Desktop App behavior.

Close Codex Desktop App and all its processes in Task Manager. Then:

1. Edit:

C:\Users\<username>\.codex\config.toml

1. Remove the following lines:

model_catalog_json = "C:\\Users\\<username>\\.codex\\ollama-launch-models.json"
model = "qwen3.6:35b"

1. Delete:

C:\Users\<username>\.codex\ollama-launch-models.json

1. Restart Codex Desktop App.

After performing these steps:

• Native OpenAI models returned.
• GPT-5.5 became selectable again.
• Codex Desktop App behaved normally.
• The configuration changes did not reappear after multiple restarts.