ISSUES HAVE BEEN GOING ON A LONG TIME
What version of Codex is running?
thread ID 019aac7e-6909-78d2-afb2-84b2e1321170
What subscription do you have?
Business
Which model were you using?
ChatGPT-5
What platform is your computer?
Unbuntu 24
What issue are you seeing?
I want to make this extremely clear: I’m dealing with two separate issues — one with the cloud AI model and one with the Codex terminal model — and both problems have been going on for a long time. I’ve been putting up with them, but I’m at the point where I need this fixed.
- Cloud Model Issues:
When I first started using the cloud model, the performance was excellent — strong intelligence, consistent reasoning, and reliable output. Over time, the performance dropped noticeably. This is not something that just happened recently. I’ve seen it decline for a while, and I kept tolerating it, but it’s undeniable at this point.
The cloud model loses track of instructions, hallucinate more often, gets confused easily, and the overall IQ feels lower than it used to be.
I shouldn’t have to repeat basic instructions over and over.
This behavior didn’t exist when I first started using your service.
- Codex Terminal Issues:
Codex has also been unreliable for a long time. There have been times when it sat for 8 hours doing nothing. It was “reading a file” and never started the task. I literally fell asleep, woke up the next morning, and it was still stuck on the same part.
Today, it spent 50 minutes just reading a file — something Claude Code does in 15 seconds.
This isn’t a new problem.
Codex used to work flawlessly when I first installed it, and now it consistently gets stuck on tasks that should take seconds, not hours.
Here’s the truth:
I already use other platforms — Grok 4.1, Claude Code, and Anthropic. Claude Code worked extremely well at first. Grok’s updates have been great. I didn’t open these accounts as an experiment — I opened them because your performance issues forced me to.
I am not a casual user. I’ve been loyal to OpenAI for a long time.
But I need reliability. I need consistency.
And the quality drop — over time — is too obvious to ignore.
If I can’t rely on your cloud model or Codex to perform the way they used to, I will move my work to the platforms that are delivering the performance I need. I’ve already started preparing for that move because these issues have gone on long enough.
This isn’t a threat — it’s honest feedback.
I want to stay with OpenAI, but I need the level of performance I originally signed up for.
Please fix whatever is affecting my account so I can continue investing in your ecosystem.
Thank you.
What steps can reproduce the bug?
Uploaded thread: 019aac7e-6909-78d2-afb2-84b2e1321170
What is the expected behavior?
_No response_
Additional information
_No response_
3 Comments
Potential duplicates detected. Please review them and close your issue if it is a duplicate.
Powered by Codex Action
Thanks for the bug report. We'll share the feedback with the team responsible for training our codex models. We're tracking this feedback and similar reports so we can make future models more robust and efficient. And thanks for using the
/feedbackcommand. That additional information is useful.Can you please share the following longer response to all departments?
Subject: Severe Degradation in Cloud Model & Codex Performance — Request for Immediate Technical Review
Hello,
I’m reaching out because the performance of both the cloud AI model on my account and Codex on my terminal has been severely degraded for a long time. I’ve been putting up with it, but at this point it’s directly interfering with my ability to work.
---
When I first started using the cloud AI on this account, the model was sharp, consistent, and felt genuinely intelligent. Over time, the quality has noticeably declined.
Here’s what I’m seeing now:
The model loses track of instructions more often
I have to repeat the same directions multiple times
It hallucinates or fabricates things more frequently
It gets mixed up in my logic or context
The overall “IQ” or reasoning ability feels significantly lower than when I started
This isn’t about one bad reply here and there — it’s a pattern.
The way the cloud model behaves today is not the way it behaved when I originally signed up. The difference is obvious from my side, and I’ve tolerated it for a long time, but it has reached a point where it’s no longer workable.
I need someone to actually look at my account and determine if something on the backend — throttling, flags, routing changes, or other internal limits — is affecting the quality or behavior of the cloud model attached to my account.
---
On top of that, Codex running on my terminal has also become extremely unreliable.
Examples:
I’ve had Codex sit for 8 hours straight “reading a file” and never actually start the task
There was a case where it effectively sat for almost 48 hours on a similar non-operation
Today, it spent 50 minutes just reading a file — something that Claude Code handles in about 15 seconds
I’m not talking about huge, exotic workloads — I’m talking about basic operations that other models complete instantly. Codex used to perform extremely well on my machine. Now it constantly gets stuck, stalls, or never begins the actual work.
Given that:
Other providers (like Claude Code and Grok 4.1) can handle similar or heavier tasks with no issue
These problems have persisted over time
And both the cloud model and Codex are showing degraded behavior
—it strongly suggests that this is not a simple configuration mistake on my side, but something account-level or system-level that needs investigation.
---
I already have active accounts with:
Anthropic / Claude Code
Grok 4.1
Other competing AI platforms
Claude Code worked extremely well from the moment I tried it. Grok’s recent update has also been very strong. I did not open those accounts casually — I opened them because I could clearly see performance on my OpenAI account getting worse over time.
I’ve been loyal to OpenAI for a long time and would prefer to keep building on your ecosystem. But if I can’t rely on either the cloud model or Codex to perform consistently and intelligently, I will have no choice but to shift my work fully to the platforms that are currently delivering the reliability and speed I need.
---
I am asking for:
I’m not asking for special treatment — just a return to the level of performance I had when I started, and a clear confirmation that nothing on your end is silently limiting or degrading my experience.
Thank you for your time and attention to this. I’m hoping this can be properly investigated and resolved so I can continue investing my time and money into OpenAI’s ecosystem.
"Success is not just about what you accomplish; it's about what you inspire others to achieve. The only way forward is together. The human heart must match the human mind."
[cid:storage_emulated_0_Android_data_com_samsung_android_email_provider_files__EmailTempImage_3_TempSignature_signature_image001_jpg_1756910082894]
[cid:storage_emulated_0_Android_data_com_samsung_android_email_provider_files__EmailTempImage_3_TempSignature_signature_image002_png_1756910099843]
Christopher Gray
Chief Executive Officer and Founder of American Blackbird. LLC.
Moble: 410-622-6620<tel:410-622-6620>
Office: 888-807-4302
Email: @.***
2118 Sunbriar Lane
Gwynn Oak MD 21207
www.amblackbird.com<http://www.amblackbird.com>
[Title: LinkedIn - Description: image of LinkedIn icon] <http://linkedin.com/> [Title: Twitter - Description: image of Twitter icon] <http://twitter.com/> [Title: google+ - Description: image of google+ icon] <http://google.com/>
This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error, please notify the sender immediately and delete this email from your system. You may not disclose, copy or distribute the contents of this email as it is strictly prohibited
-------- Original message --------
From: Eric Traut @.*>
Date: 11/22/25 3:29 PM (GMT-05:00)
To: openai/codex @.*>
Cc: Christopher Gray @.>, Author @.>
Subject: Re: [openai/codex] ISSUES HAVE BEEN GOING ON A LONG TIME (Issue #7164)
[https://avatars.githubusercontent.com/u/223546144?s=20&v=4]etraut-openai left a comment (openai/codex#7164)<https://github.com/openai/codex/issues/7164#issuecomment-3567025338>
Thanks for the bug report. We'll share the feedback with the team responsible for training our codex models. We're tracking this feedback and similar reports so we can make future models more robust and efficient. And thanks for using the /feedback command. That additional information is useful.
—
Reply to this email directly, view it on GitHub<https://github.com/openai/codex/issues/7164#issuecomment-3567025338>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BRP46IXAY2Z2ZKRWWVIG2BL36DBTBAVCNFSM6AAAAACM5ASSBSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTKNRXGAZDKMZTHA>.
You are receiving this because you authored the thread.Message ID: @.***>