codex - 💡(How to fix) Fix GPT-5.5 Fast suddenly feels as slow as Standard, with long thinking/context/search stalls

Summary

Over the past two days, GPT-5.5 Fast in Codex has felt significantly slower than before. It now feels closer to the usual Standard speed rather than Fast.

Some simple tasks that would normally complete quickly are now taking 10–20+ minutes, and sometimes even longer, before the task finishes. This feels like a clear regression compared with the previous normal behavior.

What I am observing

• GPT-5.5 Fast feels noticeably slower than it did before.
• Simple tasks can take 10–20+ minutes or longer to complete.
• The “thinking” phase can get stuck for 40+ seconds, sometimes around a minute or more.
• Automatic context compression also appears to take a very long time.
• Reading and searching steps are also extremely slow.
• At the same time, GPT-5.5 Fast also seems less capable / less reliable than before, so the experience feels both slower and less effective.

Why I do not think this is just my local machine or network

I do not believe this is caused by my computer performance. I am using a high-end MacBook Pro configuration:

• MacBook Pro 16-inch
• Apple M5 Max
• 128 GB memory
• macOS Tahoe 26.5
• Codex version 26.519.41501 (3044)

I also do not believe this is simply a network speed issue, because the slowdown is specifically noticeable in Codex/GPT-5.5 Fast behavior and is very different from how it normally performed before.

Expected behavior

GPT-5.5 Fast should feel meaningfully faster than Standard, especially on simple tasks. Thinking, context compression, reading, and searching should not regularly stall for long periods on basic requests.

Actual behavior

GPT-5.5 Fast currently feels like Standard speed or worse, with long stalls during thinking, automatic context compression, reading, and searching.

Request

Could the team please check whether there has been a recent performance regression, routing change, capacity issue, or other backend/client issue affecting GPT-5.5 Fast in Codex?

I am also curious whether other users are seeing the same thing, because this slowdown feels very obvious compared with the previous normal GPT-5.5 Fast experience.