codex - 💡(How to fix) Fix GPT-5.5 Fast mode feels like Standard speed while using increased plan usage [1 comments, 2 participants]

Summary

I am seeing a noticeable speed degradation with GPT-5.5 in Codex even though Fast mode is enabled.

Earlier today, around 3 PM ET when GPT-5.5 first became available for me, Fast mode felt genuinely fast and clearly matched the expected Fast experience. However, later in the evening around 7-8 PM ET, after more users appeared to receive access to GPT-5.5, the speed became noticeably slower.

It is not completely unusable, but it no longer feels like the Fast speed I experienced earlier today. It feels much closer to Standard speed.

Environment

• Product: Codex desktop app
• App version: 26.422.21459 (2049)
• Model: GPT-5.5
• Intelligence level: Extra High
• Speed setting: Fast
• Approximate time observed: April 23, 2026, around 7-8 PM ET
• Comparison point: April 23, 2026, around 3 PM ET, shortly after GPT-5.5 became available

What happened

Fast mode is enabled, so usage is being charged as Fast / increased plan usage. However, the actual experience feels much closer to Standard speed.

As one recent example, I ran a very simple task and it took about 7 minutes 49 seconds. Based on how GPT-5.5 Fast behaved earlier today, I would have expected a small task like this to finish in roughly 5-6 minutes or less.

Expected behavior

If Fast mode is enabled and charged with increased plan usage, the actual response speed should remain consistently faster than Standard mode.

If server load is too high to maintain stable Fast performance, I think the Fast usage cost should be reduced, paused, or adjusted accordingly. Otherwise, users are paying the Fast usage cost while effectively receiving Standard-like speed, which feels unfair.

Request

Could the team please investigate whether GPT-5.5 Fast mode is being degraded under higher server load or increased rollout traffic?

I would appreciate it if the team could seriously look into this and consider a fair handling of Fast usage charging when the backend cannot consistently provide Fast-level performance.

<img width="1116" height="748" alt="Image" src="https://github.com/user-attachments/assets/59145254-b1dc-4f5e-b786-9270828d1a57" /> <img width="1694" height="262" alt="Image" src="https://github.com/user-attachments/assets/10f03599-1048-4edd-9fbc-94e785ea4745" />