Codex 5.3 Disconnecting and Running Extremely Slowly

[jonlave]

What version of the Codex App are you using (From “About Codex” dialog)?

MacOS Version 260208.1016 (571)

What subscription do you have?

Pro

What issue are you seeing?

Was running Codex 5.3 on the Mac OS app for the last 3 days without issue. Set the thinking setting to Extremly High, still no problems, but as of around 5:00 PM PST today (Feb 9), all of a sudden, the thread hangs with no updates on the page. I left it like that for an hour and came back and still no updates/messages on the screen so I quite the app and restarted. Prompted it to "retry" and it still hangs on "Thinking", showing messages like "Reconnecting 1/5". Eventually, it did start showing messages, but it's much, much slower than before. I have a skill to just clean up the local git state and switch back to main. This was consistently taking about 10s or less. Now It takes almost 8 minutes. Any ideas what happened tonight?

What steps can reproduce the bug?

Running the Codex app on Mac.

What is the expected behavior?

It should be thinking and working 10x faster than it is.

Additional information

No response

[etraut-openai]

Thanks for the bug report. If you're able to repro this, please use the /feedback slash command to upload your logs and session details, then post the thread ID here. I can take a look at the data and see if I can diagnose.

[jskorlol]

Since the release of the Codex 5.3 API, this issue has started to occur.
At the very least, I hope that users of the Codex CLI and app are not experiencing slower performance.
What’s most concerning is that the performance is actually worse than Codex 5.2.

[etraut-openai]

@jskorlol, if you're seeing slow performance and think something is wrong, please use /feedback to upload your logs and session details and post your thread ID.

[jskorlol]

@etraut-openai
019c45f6-20eb-7251-bc16-20f18f4a76fe
019c45bb-f366-7b72-9d00-1dce6f0d7df5

It’s become slower and the performance is worse than it was in 5.2.

[jonlave]

@etraut-openai - here's my feedback id's. Thanks

019c452d-5845-7491-ab3b-cb33cc71610a
019c451a-ff3e-7cf0-85ae-a2ab496d8228

[pbowyer]

And mine: 019c4758-d74a-7df3-bda4-524283b1e298

[christian-wiloejo]

@etraut-openai Here's my feedback ID: 019bfb21-b535-71a3-94fa-ee279c55a1b5

[marjan2k]

@etraut-openai some of my feedback IDs:

• 019c47ca-763c-7272-8835-847f56ca81e1
• 019bcd98-f494-7762-8d14-36b35a2df492

[IlyaShorin]

same here 019c47ce-36ff-7e21-a537-ac6544097883

[jonlave]

Adding some screenshots of what I'm seeing. A simple prompt to merge a PR and clean the local git state was taking about 10 seconds before last night, but now is in the order of several minutes. When asked to investigate why it's taking that long, Codex claims it's not taking that long and that there is either a lot of latency in the chat or network delays. I'm on gig speed internet, in a metro area and it's been running fine. I would like the app to record the chat latency so we can get a better sense of where the time is being spent.

[MisterRound]

Likely related to #11189 assume you're all actually seeing redirected 5.2 responses

[jonlave]

It's back to normal now. The other thread that @MisterRound linked (#11189) basically identifies the culprit which is OpenAI flagging our accounts as potentially be using the models for cyber attacks so we have to submit government issued IDs to gain back access. Not sure why I got flagged. I was using it heavily for a few days to work on my B2C SaaS.

[Karimrekiki]

The applicatioin is unusable on my m2 ultra 64gb. Switching to different threads takes ages. it runs very buggy. the AI part is super quick however the interface is close to useless. Losing my nerf here. when you submit a question you cannot be sure it was actually submitted.