GPT-5.5 Fast is severely throttled: slower than 1/20 of Standard even after multiple app restarts
Codex app version
Desktop Codex: 26.608.12217 — released Jun 8, 2026
Platform / plan
- macOS Tahoe
26.5.1 - Apple Silicon MacBook Pro
- Pro plan
- Mode: GPT-5.5 Fast
Related feedback ID
019eb48b-00d0-7b91-b50a-cbd119bb5c4cSummary
GPT-5.5 Fast is currently experiencing a severe speed degradation. The actual response and task progress speed feels slower than 1/20 of Standard mode, even though Fast mode consumes higher usage.
I have restarted the Codex app multiple times, but the speed did not improve at all. This does not look like a temporary UI glitch or local app state issue.
Observed behavior
- GPT-5.5 Fast is extremely slow.
- The speed is far below what Fast mode should provide.
- In practice, it feels slower than even 1/20 of Standard mode.
- Restarting the Codex app multiple times does not fix the issue.
- The degraded speed makes normal task execution extremely inefficient.
Expected behavior
GPT-5.5 Fast should be meaningfully faster than Standard mode, especially because it consumes higher usage. If Fast mode is degraded, throttled, or affected by backend capacity issues, the product should clearly communicate that to users.
Actual behavior
Fast mode currently consumes Fast-mode usage but delivers a speed that feels dramatically slower than Standard. The issue persists even after restarting Codex multiple times.
Why this matters
This is a serious usability and fairness problem. Users choose Fast mode because they need faster task completion and accept the higher usage cost. If Fast mode becomes slower than Standard while still charging higher usage, the weekly limit is wasted very quickly without receiving the expected performance.
For Pro users, this makes the 20x weekly usage limit much less useful, because tasks consume expensive Fast-mode quota while progressing extremely slowly.
Request
Please urgently investigate whether GPT-5.5 Fast is currently affected by throttling, backend latency, routing issues, or a performance regression.
If Fast mode is temporarily degraded, please make that visible in the UI and consider crediting back the wasted Fast-mode usage for affected users.
https://github.com/user-attachments/assets/cddebe8a-1442-4624-9463-0b8ff7ea8746
4 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
From what I can tell this isn't tied to fast mode, I think gpt 5.5 is having an outage? Not sure - status page seems fine. 5.4 seems to be working normally
@etraut-openai I honestly cannot believe this. This is a very simple task that would normally take about 20 minutes to finish, but it has already been running for nearly five hours. The streaming speed has not improved at all. It is still painfully slow.
At the current extremely slow pace, this task will probably take more than 8 hours to complete. That would shamefully break the longest-task record on my account. This is not something to be proud of. It is not because more work is being done over a longer period of time; it is simply because the speed has become unbelievably slow.
I am also amazed by how slowly the Codex team is handling this issue. It feels like these problems are being ignored. I already reported this problem around noon yesterday, but by early morning today it still has not been fixed, and nobody has explained the situation to me. Is this really how the Codex team handles serious user-facing problems?
Please check the backend yourselves. Look at how many tokens this task has consumed and you will understand how slow this task actually is.
This has wasted an entire day of my time. A task that should have taken about 20 minutes has turned into something that takes the whole day. Honestly, if I had done this manually myself, I could probably have finished it in only a few hours.
I do not know what the team thinks when seeing this problem right now. I just want to say: if Codex continues like this, more users will move to other platforms.
Related feedback ID:
019eb48b-00d0-7b91-b50a-cbd119bb5c4c
<img width="1318" height="184" alt="Image" src="https://github.com/user-attachments/assets/88b4662b-4576-4561-8bf2-535167266ec9" />
https://github.com/user-attachments/assets/19bc0a90-f609-41e1-bba2-2d263e84449d
Confirming the severe 2026-06-11 speed degradation from another high-tier / 20x account.
Environment observed:
codex-cli 0.139.06.6.87.2-microsoft-standard-WSL2Symptoms:
This matches the current-day degradation angle here. From the user side it feels like paid high-tier capacity is not translating into usable response speed today.