Codex Desktop auto-selects gpt-5.2-codex for ChatGPT Plus accounts and every new thread fails

What version of the Codex App are you using?

26.519.81530 (3178)

What subscription do you have?

ChatGPT Plus

What platform is your computer?

macOS 26.4.1 (Apple Silicon)

What issue are you seeing?

Every new Codex Desktop thread is being created with gpt-5.2-codex as the active model, and requests immediately fail with:

{"type":"error","status":400,"error":{"type":"invalid_request_error","message":"The 'gpt-5.2-codex' model is not supported when using Codex with a ChatGPT account."}}

This happens even though:

• the app is fully updated
• I create completely new threads
• I reset local Codex caches/storage/state
• I changed local global agent mode from custom to auto

Steps to reproduce

1. Open Codex Desktop on macOS
2. Create a brand new thread
3. Do not manually choose a model
4. Send any prompt
5. The thread fails immediately with the unsupported-model 400 error

What is the expected behavior?

New threads should default to a model actually supported for ChatGPT accounts, or unsupported models should never be auto-selected.

What have you already tried?

• App restart
• New threads
• Local state DB cleanup
• Local Storage / Session Storage / SharedStorage reset
• Global state reset from agent-mode=custom to agent-mode=auto

Additional evidence

A newly created session still records turn_context.model = "gpt-5.2-codex" on the latest app/CLI pipeline.

Relevant local evidence:

• app version: 26.519.81530 (3178)
• session cli_version: 0.133.0
• new session turn_context.model: gpt-5.2-codex

I can provide the session JSONL snippet if needed.