GPT-5.5 Fast mode is currently much slower than Standard mode

Resolved 💬 1 comment Opened May 27, 2026 by GGBondBlueWhale Closed May 27, 2026

I am a $200/month Pro user.

Over the last few days, GPT-5.5 Fast mode in Codex has become extremely slow and has stayed that way for about four days. I kept using Fast because it is supposed to be the faster option, but just now I accidentally discovered that switching to Standard mode makes responses much faster — several times faster than Fast in my real usage.

This is very frustrating because Fast mode has been painfully slow for days, while still consuming my weekly usage. My current weekly limit is now down to about 9%, largely because I continued using the extremely slow Fast mode before realizing Standard was actually faster.

Expected behavior:

  • Fast mode should be clearly faster than Standard mode.
  • If Fast capacity is degraded, the UI should warn users or automatically avoid charging/using extra Fast-mode usage.
  • Users should not lose a large amount of weekly usage because the advertised faster mode is slower than Standard.

Please investigate and fix the Fast-mode performance regression as soon as possible. I also strongly suggest compensating affected users by restoring some weekly usage / limit, because this is unfair to users who spent the last several days using the broken Fast mode.

View original on GitHub ↗

This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗