I’m experiencing persistent Codex failures on both the VS Code extension and Codex CLI.
Environment:
Subscription: ChatGPT Plus
OS: Windows
Codex CLI: v0.163.0
Models tested: GPT-6 Astra Extra High and GPT-5.6 Sol Medium
5-hour allowance remaining: 77%
Weekly allowance remaining: 96%
Date: October 20, 2026 (UTC+7)
Error: Selected model is at capacity. Please try a different model.
The same error happens even when submitting a simple prompt such as “Hello, reply OK only” in a new session.
I have tested both the VS Code extension and standalone CLI, and both fail.
Similar symptoms have been reported in GitHub issue #53569.
Could you please investigate whether my account is affected by a model-serving capacity, admission-control, or account-routing issue? My available usage allowance suggests this is not a normal quota exhaustion problem.
I can provide screenshots and additional diagnostics privately if required.
As of yesterday (20th October) i have been getting very very very frequent “Selected model is at capacity. Please try a different model.” responses. ChatGPT / Codex is practically unusable for me at the moment. About 90% of my messages end up with that response. How is this still an issue?
I encountered the exact same “CAPACITY” problem, very serious from 10 days ago. It eased up a week ago and improved significantly—but now the issue has started all over again!
What version of the Codex App are you using (From “About Codex” dialog)?
26.1016.51333
What subscription do you have?
pro 5x
What platform is your computer?
Microsoft Windows NT 10.0.26200.0 x64
What issue are you seeing?
Selected model is at capacity. Please try a different model.
What steps can reproduce the bug?
Feedback ID: 01a1424d-18d9-71a3-8aff-defdd6a1f867
What is the expected behavior?
Selected model is at capacity. Please try a different model.
It hasn’t been working properly since the last reset—it’s already been two days.
Additional information
Please resolve this issue as soon as possible, or provide a clear explanation for the error. I have paid for this service, so I am entitled to use it and to know the reason for the problem.
caramba @OpenAI_Support resolvam esse problema meu, muitos usuarios reclamando e vocês pouco se importando com os usuarios, ta prejudicando uso e consumindo os tokens pagos pelos usuarios, se o problema é capacidade porque não colocam mais servidores(capacidade) se é o problema?
resolvam, a gente usuário paga caro pelo serviço pra usar AI!
I’m experiencing a persistent issue with ChatGPT Work/Codex on Windows 11. The application has become practically unusable, even though I have available usage limits on my ChatGPT Plus subscription.
Errors encountered
“Selected model is at capacity. Please try a different model.”
“Optimization interrupted”
“Error running remote compact task”
These errors occur across different models, new conversations, and even simple prompts like “Hi”. Occasionally, one or two prompts work before the errors return.
Troubleshooting already completed
Reinstalled the application.
Tested multiple models and new conversations.
Changed IPv4 DNS to Cloudflare.
Configured IPv6 DNS.
Temporarily disabled IPv6.
Flushed the DNS cache.
Tested DNS resolution and TCP connectivity to codex-cloud-backend.chatgpt.com on port 443.
Tested HTTPS over both IPv4 and IPv6.
Tested Proton VPN on the laptop.
None of these steps resolved the issue on my home Wi-Fi.
Important observation
ChatGPT Work works on my mobile hotspot but consistently fails on my home Wi-Fi.
Both connections can reach the backend. A direct curl request receives a Cloudflare 403 challenge on both networks, so that response alone does not explain the difference.
Regular ChatGPT in Chrome also works.
Environment
OS: Windows 11
Application: ChatGPT Work / Codex desktop
Subscription: ChatGPT Plus
Issue observed: October 2026
Region: India
Usage quota: Available, not exhausted
Support status
I’ve already contacted OpenAI Support under case #17857627 and provided diagnostic logs.
Questions
Has anyone experienced the same issue, particularly where Work functions on a mobile hotspot but not home Wi-Fi?
Could this be related to WebSocket connections, ISP routing, or backend session handling?
Is there a known fix or ongoing investigation?
I’d appreciate any guidance from the community or OpenAI engineering team. The application is currently unreliable for everyday work.
It’s affecting the API, Codex at night—everything. User applications that rely on integrations are failing. I’m currently compiling a list of all the affected users so we can report it; it wouldn’t make sense for just one person to reach out. These users pay a high price for the service, yet when they need to use it, no fixes are being implemented. We need to know your stance on this error.
It’s becoming really frustrating. The only reason I haven’t switched to another service is that I simply can’t—it’s not an option. Beyond that, you’re letting your users down. I already lost one workspace, and now that I’ve paid for a new one, I’m running into the same problem again.
