All models — Codex CLI hangs indefinitely on all prompts, no response generated

[galcianuk]

What version of Codex CLI is running?

v0.111.0, v0.112.0

What subscription do you have?

Plus

Which model were you using?

gpt-5.4, gpt-5.3

What platform is your computer?

Windows

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

WSL

What issue are you seeing?

Codex CLI accepts prompts but produces massively delayed response. The model appears to hang indefinitely — no streaming output, no error, no timeout. This
affects all prompt types (simple greetings, questions, codebase analysis). The status bar shows "gpt-5.4 high · 100% left" throughout, suggesting no tokens are being consumed.

What steps can reproduce the bug?

Steps to reproduce:

1. Launch codex (v0.112.0) with model gpt-5.4 high
2. Enter any prompt (e.g., "Hello", "Explain this codebase")
3. Wait — no response is ever generated

What is the expected behavior?

Model should respond.

Additional information

Observed behavior:

• Prompt is accepted and displayed in the conversation
• No streaming output begins
• No error message or timeout
• Status bar remains at "100% left"
• Tested with 5 consecutive prompts over several minutes — received a response eventually after 5 minutes.

[github-actions]

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

• Codex CLI TUI accepts prompt but renders no assistant response on Linux (exec works) #13726
• codex hanging for a couple minutes after prompt #13774
• macOS local Codex clients hang after sending prompts #13715

Powered by Codex Action

[etraut-openai]

It sounds like you suspect this is a regression? Let's see if we can confirm that because it will inform prioritization. If you downgrade to 0.111.0, does the problem go away?

[galcianuk]

No, it doesn't go away after updating. And I also noticed it happens on 5.3 as well.

[etraut-openai]

OK, then this isn't a regression. I don't think we've received other reports of this, so it's possible that this is specific to your system. Perhaps you changed something recently in your shell configuration or Codex configuration? Or maybe there's something corrupt within your .codex directory.

[zouyang666]

I have the same problem.

[galcianuk]

OK, then this isn't a regression. I don't think we've received other reports of this, so it's possible that this is specific to your system. Perhaps you changed something recently in your shell configuration or Codex configuration? Or maybe there's something corrupt within your .codex directory.

Nothing has changed. It stopped working after the release of gpt 5.4 :(.

[nilswx]

Same problem (gpt-5.4 and gpt-5.3-codex), happens predominantly right around forced compaction time (10-20%). I have to remind myself to /compact or risk being stuck for 30+ minutes.

'Remote compaction task' also fails more often. Problem seems serverside (session cache invalidation?). I'm able to get a quick response in a new chat, but even codex fork on the old (long) chat leads to a long (and quiet) stall.

codex-cli 0.113.0-alpha.1 and macOS app Version 26.305.950 (863) (codex-cli 0.108.0-alpha.12)

EDIT: still happens after rotating VPN IP and cp auth.json dance to switch ChatGPT accounts. After around 15 minutes it timed out with Error running remote compact task: stream disconnected before completion: error sending request for url (https://chatgpt.com/backend-api/codex/responses/compact), seems unrelated to server caches. Quiet compaction overrun, buffer too small?

[shivakanthsujit]

Yes I am having the same issue. Started right after I updated to the newest version 10 minutes ago. Was working before the update. After I send a message I dont get any response, and the cli just hangs.

[songkeys]

Having the same issue. 10 minutes ago.

v0.112.0. It worked fine ten minutes ago in the same version, but suddenly it just hangs on every message.

[puzanov]

I have the same issue which started to happen today after I've updated to v0.112.0.
Box: Arch Linux
Behavior: Worked well after update for a hour or two, right now all prompts do not have any responses.
Even the command /exit does not work so I have to kill codex from the process list.

[songkeys]

RUST_LOG='codex_api::sse::responses=trace' codex exec --sandbox read-only --model gpt-5.3-codex 'ping'
OpenAI Codex v0.112.0 (research preview)
--------
workdir: ...
model: gpt-5.3-codex
provider: openai
approval: never
sandbox: read-only
reasoning effort: xhigh
reasoning summaries: none
session id: 019cd167-d5c3-7691-a3af-ce39bd4ca827
--------
user
ping
2026-03-09T07:02:56.577787Z TRACE codex_api::sse::responses: unhandled responses event: response.in_progress
2026-03-09T07:02:58.437882Z TRACE codex_api::sse::responses: unhandled responses event: response.in_progress
2026-03-09T07:03:02.608218Z TRACE codex_api::sse::responses: unhandled responses event: response.content_part.added
2026-03-09T07:03:02.708617Z TRACE codex_api::sse::responses: unhandled responses event: response.output_text.done
2026-03-09T07:03:02.708654Z TRACE codex_api::sse::responses: unhandled responses event: response.content_part.done
codex
pong
tokens used
17,582

I think I just found the issue. It looks like the response is not handled at all; only the result is printed.