Codex CLI 0.66.0 and Azure GPT-5 - 429 Too Many Requests
Resolved 💬 5 comments Opened Dec 10, 2025 by anitagirme Closed Mar 29, 2026
💡 Likely answer: A maintainer (etraut-openai, contributor)
responded on this thread — see the highlighted reply below.
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 execwith --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_
5 Comments
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 Requestserror when I switch back to CLI0.63.0version. 🤔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.
@etraut-openai No longer seeing the
429 Too Many Requestserror with the latest version after increasing the TPM.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.