Using service_tier = "flex" in CLI produces error 400 on message

Open 💬 4 comments Opened Mar 18, 2026 by Pimpmuckl

💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What version of Codex CLI is running?

0.115.0

What subscription do you have?

Pro

Which model were you using?

gpt-5.4

What platform is your computer?

Microsoft Windows NT 10.0.26200.0 x64

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

Windows Terminal PS7

What issue are you seeing?

Error {"type":"error","status":400,"error":{"type":"invalid_request_error","message":"Unsupported service_tier: flex"}} when sending messages or establishing websocket with service_tier = "flex"

What steps can reproduce the bug?

set service_tier = "flex" in config.toml as per official OpenAI Codex documention
start codex-cli
send agent message

What is the expected behavior?

No Error 400 and agent response

Additional information

Same behaviour in Codex App, though it falls back to "normal" so this is masked to the user.

View original on GitHub ↗

4 Comments

github-actions[bot] contributor · 5 months ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

#14204

Powered by Codex Action

etraut-openai contributor · 5 months ago

Are you using an Azure-hosted endpoint or OpenAI?

Please use /feedback to upload your logs and post the thread ID here.

Pimpmuckl · 5 months ago

Are you using an Azure-hosted endpoint or OpenAI? Please use /feedback to upload your logs and post the thread ID here.

I assume the answer is OpenAI as that's default, I haven't changed anything about endpoints. Physical location is in Cologne, Germany, unsure if there is opportunistic routing to different providers depending on location.

/feedback thread: 019d02e1-a69f-7e73-906e-6f3debc5d03f

Mygod · 4 months ago

Also reproducing on Linux.