[macOS] Severe performance regression: 100%+ CPU, 10+ GB RAM usage, and frequent UI hangs

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

26.810.41047 (6570)

What subscription do you have?

ChatGPT Plus

What platform is your computer?

macOS 26.6.1 (Build 25G76), Apple Silicon arm64, Mac15,7, 36 GB RAM

What issue are you seeing?

The macOS Codex app currently has a severe performance problem and has become extremely frustrating to use.

This is not a minor UI slowdown.

On my machine, the Codex/ChatGPT process frequently reaches:

• 100%+ CPU usage (104.9% observed)
• More than 10 GB of memory usage (10.4 GB observed)
• Long periods of UI freezing / complete unresponsiveness
• Severe input lag, scrolling lag, and delayed interactions

macOS has also generated a hang report for the application. In the attached report, the process was unresponsive for approximately 22 seconds before sampling.

System information:

• Codex version: 26.810.41047 (6570)
• macOS: 26.6.1 (Build 25G76)
• Architecture: arm64
• Hardware: Mac15,7
• RAM: 36 GB
• CPU cores: 12

A desktop application consuming over 10 GB of RAM while also sustaining 100%+ CPU usage during normal usage is not reasonable, especially on a machine with this hardware.

The performance of the Codex desktop app has noticeably degraded across recent versions. Normal operations such as typing, switching conversations, scrolling through long sessions, and interacting with the UI can become extremely sluggish, and sometimes the entire application stops responding.

Please investigate this as a performance regression rather than treating it as a generic "app feels slow" report.

In particular, please investigate:

1. CPU usage during idle and active conversations
2. Memory growth during long-running sessions
3. Possible memory leaks
4. Rendering performance with long conversations
5. Background processing that continues consuming CPU
6. UI/main-thread blocking
7. Whether conversation/context size causes uncontrolled resource growth

I have attached screenshots showing the CPU and memory consumption, as well as the macOS hang report.

At this point the performance degradation is seriously affecting daily use of Codex. I would strongly recommend profiling the macOS desktop app under long-running, real-world sessions instead of only testing fresh or short conversations.

Please prioritize this. The current level of resource consumption and UI freezing is not acceptable for a production desktop application.

What steps can reproduce the bug?

1. Launch the Codex desktop app on macOS.

2. Use Codex normally for an extended period of time, including:

   • Sending messages in existing conversations
   • Working with longer conversations
   • Scrolling through conversation history
   • Switching between conversations
   • Continuing to interact with the app without restarting it

3. Monitor the Codex/ChatGPT process in Activity Monitor.

4. Over time, memory usage increases significantly. In my case, the process reached approximately 10.4 GB of RAM.

5. CPU usage can also remain extremely high. I observed the process reaching approximately 104.9% CPU.

6. As resource usage increases, the UI becomes progressively slower:

   • Typing starts to lag
   • Scrolling becomes sluggish
   • Switching conversations becomes delayed
   • UI interactions stop responding promptly

7. Eventually, the application may completely hang for an extended period. macOS recorded one instance where the process was unresponsive for approximately 22 seconds before the system began sampling it.

No unusual workload or special reproduction setup is required. This happens during normal day-to-day usage of the Codex desktop app.

The issue appears to become significantly worse during long-running sessions and when working with larger/longer conversations.

I have attached Activity Monitor screenshots and the macOS hang report showing the observed CPU usage, memory consumption, and application hang.

What is the expected behavior?

No response

Additional information

No response

Start by reproducing the issue in the macOS Codex app during extended use of long conversations, while monitoring the Codex/ChatGPT process in Activity Monitor and reviewing the attached macOS hang report. Compare CPU, memory, typing, scrolling, conversation switching, and UI responsiveness over time; done means the regression and resulting hangs are identified and normal usage no longer produces the reported resource growth or freezes.