Simply garbage, the model's intelligence has severely degraded.
Open 💬 2 comments Opened Jun 2, 2026 by s7monk
💡 Likely answer: A maintainer (github-actions[bot], contributor)
responded on this thread — see the highlighted reply below.
What version of the Codex App are you using (From “About Codex” dialog)?
26.527.60818
What subscription do you have?
pro
What platform is your computer?
win11
What issue are you seeing?
The model has suffered a severe decline in intelligence. Currently, the GPT-5.5 xhigh model has experienced a very serious drop in intelligence, and its actual capability is extremely poor, to the point of being unbearable. Its real performance is definitely inferior to DeepSeek V4, and its performance is very bad.
What steps can reproduce the bug?
The model has suffered a severe decline in intelligence. Currently, the GPT-5.5 xhigh model has experienced a very serious drop in intelligence, and its actual capability is extremely poor, to the point of being unbearable. Its real performance is definitely inferior to DeepSeek V4, and its performance is very bad.
What is the expected behavior?
_No response_
Additional information
_No response_
2 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
From my personal experience, this issue started around May 23.
I cannot prove what is happening internally, but it feels like whenever a new model is about to be released, the older model suddenly becomes slower and less capable, so users can feel a much bigger gap between the old model and the new one. Looking back, this pattern has seemed to happen before previous major model releases as well.
To be honest, this feels very unfair to paying users. It also feels like the Codex team expects users to simply get used to the slower and weaker experience over time.
This comment is based entirely on my own experience and the experience of other developers I know who also use Codex heavily. At this point, it is not just me feeling this degradation.