What have you guys been up to lately? Because Codex has gone from being kinda slow to being ridiculously slow and stupid now. :D

[sfz009900]

What have you guys been up to lately? Because Codex has gone from being kinda slow to being ridiculously slow and stupid now :D

[sfz009900]

Since the rumors started circulating a few days ago about the increased speed, my 5.2x high-speed connection has been completely unreliable. I haven't been able to get a single satisfactory result, and I'm really frustrated :(

[etraut-openai]

Sorry to hear that that you're seeing degraded performance. I haven't observed an uptick in reports of performance issues from other users, and I don't think that Codex team members have seen any degradations lately (I certainly haven't, and I use Codex all day long).

Perhaps you changed a configuration setting without realizing it? For example, did you switch to "xhigh" reasoning mode? We generally recommend "medium" for day-to-day coding, and "xhigh" will definitely lead to more token usage and longer turn times. I recommend looking at your config.toml file and verifying that you haven't made any inadvertent changes.

[ambrose2025]

Sorry to hear that that you're seeing degraded performance. I haven't observed an uptick in reports of performance issues from other users, and I don't think that Codex team members have seen any degradations lately (I certainly haven't, and I use Codex all day long).

Perhaps you changed a configuration setting without realizing it? For example, did you switch to "xhigh" reasoning mode? We generally recommend "medium" for day-to-day coding, and "xhigh" will definitely lead to more token usage and longer turn times. I recommend looking at your config.toml file and verifying that you haven't made any inadvertent changes.

Like 8 hours ago, 5.2 xhigh has been even worse than 4o-mini...
It's completely ignoring rules and not following my prompts at all (context isn't bloated, and starting fresh conversations doesn't fix it either

[ZaiyunLin]

extremely slow even when thinking is set to minimal (gpt5.2).

[ZaiyunLin]

average around 20 minutes to get result

[yherder]

Same here. Sometimes 20-30 minutes for 200 character changed....

[alc1986]

Same issue here. Using Codex via MCP is even worse. It's taking well over an hour to get even basic responses.

[mikesibiu]

Codex was great, but definitely something changed. For me, I noticed after an update a few days ago. It is like someone gave it a lobotomy. It tells me it updated a bash script, but it never does. Then it tells me to update it and gives me instructions on what lines to change. I am cancelling my subscription and just buying an extra Claude code professional subscription instead. This is worse than useless and is costing me time and increasing my blood pressure.

[etraut-openai]

@mikesibiu, if you're seeing a problem that you think is a bug, please report it in the issue tracker. Use the /feedback slash command to upload your logs and session details, and include the thread ID in the bug report. That will allow us to investigate.

[mikesibiu]

I switched the model from 5-3 to 5-2, and I think it's back to working again. 5-3 is a gift from hell...

[mikesibiu]

I will do a /feedback soon