GPT-5.4 Fast and GPT-5.5 Fast are extremely slow (~1 token/s)

What version of the Codex App are you using (From “About Codex” dialog)?

26.519.41501 (3044)

What subscription do you have?

chatgpt pro 20x

What platform is your computer?

Darwin 25.3.0 arm64 arm

What issue are you seeing?

GPT-5.4 Fast and GPT-5.5 Fast have become extremely slow over the past 2–3 days.

Previously, responses were near real-time, but now both models frequently:

• stall during thinking
• take 10–30+ seconds before first token
• stream at ~1 token/s or even slower
• feel almost identical to standard/deep reasoning latency

This happens even with:

• low effort
• fast mode
• small prompts
• new sessions
• simple repositories/tasks

The issue became very noticeable around May 24–26.

Environment:

• Codex app
• macOS
• GPT-5.5 Fast
• GPT-5.4 Fast
• low effort mode enabled

Current experience is significantly worse than last week.

What steps can reproduce the bug?

1. Open Codex.
2. Start a new session.
3. Select GPT-5.5 Fast or GPT-5.4.
4. Set reasoning/effort to Low.
5. Send a simple coding prompt, for example:

Please explain what this function does and suggest one small improvement.

or:

Please update this component to change the button label from "Submit" to "Save".

1. Observe the response speed.

Expected behavior:
The model should respond quickly in Fast / Low mode, with normal streaming speed.

Actual behavior:
Both GPT-5.4 and GPT-5.5 Fast are extremely slow. The model often waits a long time before the first token, then streams at around 1 token/s or slower. This happens even for simple prompts and new sessions.

This started becoming very noticeable around May 24–26, 2026.

Session ID:
N/A / I can provide one if needed.

Token limit usage:
Not near the limit.

Context window usage:
Low. This also happens in new sessions with simple prompts.

What is the expected behavior?

_No response_

Additional information

_No response_