Codex CLI 0.66.0 and Azure GPT-5 - 429 Too Many Requests

[anitagirme]

What version of Codex is running?

0.66.0

What subscription do you have?

Azure

Which model were you using?

gpt-5

What platform is your computer?

No response

What issue are you seeing?

Codex retries internally, then fails with exceeded retry limit, last status: 429 Too Many Requests

What steps can reproduce the bug?

• Create Azure OpenAI GPT-5 deployment
• Configure Codex CLI to use that deployment. Use codex exec with --config parameters to configure Azure OpenAI deployment

What is the expected behavior?

This was working earlier with Codex CLI 0.63.0 and prior versions and suddenly started failing after that.

Additional information

No response

[etraut-openai]

This typically means that you are exceeding your TPM (tokens per minute) rate limit in Azure. You or your admin should be able to increase the TPM for your Azure account.

[anitagirme]

This typically means that you are exceeding your TPM (tokens per minute) rate limit in Azure. You or your admin should be able to increase the TPM for your Azure account.

Sure! I can take a look at it. But surprisingly it works without throwing 429 Too Many Requests error when I switch back to CLI 0.63.0 version. 🤔

[etraut-openai]

Interesting. It's possible that the timing has changed. We're continually looking for performance improvements in codex. It's possible that local tool invocation takes less time with the latest version, which would lead to a shorter turnaround time between model invocations.

[anitagirme]

@etraut-openai No longer seeing the 429 Too Many Requests error with the latest version after increasing the TPM.

[etraut-openai]

OK, good to know. Let's keep the issue open for now because I think we should provide a better error message in this case.