Inference routing failure

Resolved 💬 3 comments Opened Apr 3, 2026 by K4F7 Closed Apr 3, 2026

What version of Codex CLI is running?

codex-cli 0.118.0

What subscription do you have?

plus

Which model were you using?

gpt-5.4

What platform is your computer?

Microsoft Windows NT 10.0.29558.0 x64

What terminal emulator and version are you using (if applicable)?

PowerShell

What issue are you seeing?

Windows Codex CLI returns 503:
No available channel for model gpt-5.4

WSL works fine with same account.
Windows config seems pointing to local distributor (127.0.0.1:15721).

What steps can reproduce the bug?

Uploaded thread: 019d517f-86ed-7131-961a-0530c3f5b65f

What is the expected behavior?

The Codex CLI should successfully connect and return a response from the model.

Running the same setup in WSL works correctly, so the Windows CLI is expected to behave the same way without returning a 503 error.

Additional information

  • Error: 503 Service Unavailable

No available channel for model gpt-5.4
URL: http://127.0.0.1:15721/v1/responses

  • The issue only occurs on Windows Codex CLI.
  • The same account and configuration works correctly in WSL.
  • It appears that the Windows CLI is pointing to a local distributor (127.0.0.1:15721), which may not have a valid model channel configured.
  • Network setup uses TUN mode (no manual proxy). Connectivity is confirmed working.
  • Suspected mismatch between Windows and WSL Codex configuration (base_url or model settings).

View original on GitHub ↗

This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗