Skip to content

The 'deepseek-v4-flash' model is not supported - Codex

Quick fix

Enable the keep-official-login option in cc-switch, confirm local route takeover is on, then restart Codex desktop.

Error output
The 'deepseek-v4-flash' model is not supported when using Codex with a ChatGPT account.

Codex desktop (the upstream client) has a limitation: its model picker decides whether to show third-party custom models based on whether you signed in with an official ChatGPT identity. Without that identity it forcibly reverts to the official default model and blocks third-party model requests. Early versions may also have had endpoint format compatibility problems, though newer cc-switch supports the format conversion.

  1. Sign in to an official ChatGPT account once in Codex desktop, so official credentials are stored in ~/.codex/auth.json.

  2. In cc-switch settings, enable "Keep official login when switching third-party providers".

  3. Make sure local route-takeover is enabled for the third-party provider you use.

  4. Quit Codex desktop completely, restart it, and open the model picker.

ToolCodex
VersionUnknown
PlatformsWindowsmacOS
Why does the model disappear again after a while?
The preserved official credentials may have expired. Sign in to an official ChatGPT account once more in Codex to restore it.
I can see third-party models in the terminal but not in the desktop app. What now?
That means the data layer is fine and it is purely a desktop UI limitation. Sign in to the official account in the desktop app and enable the cc-switch keep-login option as described.
Can I sign in to Codex with a DeepSeek API directly?
Yes — users report signing in to Codex with a DeepSeek API directly works fine. But to relay through cc-switch, follow the steps above to preserve the official login state.

This is an unofficial community wiki with no affiliation to the cc-switch authors or the project itself. Its content is compiled from the project's public GitHub issues. This site distributes no software.