This is my second subscription, but I frequently encounter the error: “Selected model is at capacity. Please try a different model.” This prevents me from getting my work done, and the issue persists no matter how long I wait.
Because the selected model is deemed to be at capacity. ![]()
You can check system status at OpenAI here: https://status.openai.com/ - in case there is a broader issue.
Switch model or wait 30 mins.
I usually find the prompt “continue” works …
Hey @Noah15, welcome to the Community! Getting stuck on the same error after waiting is frustrating when you’ve got work to finish.
Have you tried another model, as @merefield suggested? Does that work, or do you get the same error across models, even in a new chat?
This Help Center article covers the exact message you’re seeing. A model can sometimes be unavailable even when no incident is listed on the status page.
If switching models doesn’t help either, please reach out to Support using the chat bubble at the bottom-right of help.openai.com. Please include:
- The exact error and which models you tried.
- When it last happened, including the date, time, and timezone.
- A screenshot with personal details hidden.
- Your device, browser or app version, and whether you’re using ChatGPT or Codex.
- Any request ID shown with the error, if available.
Once you have a case ID, feel free to share it here so I can take a look. If you already have a support request open for this, you can share that case ID instead.
Just make sure to share any account or billing details only in the private support chat, since this thread is public.
-Selena
Hi Selena, could you take a look at case #19012102? Persistent Codex capacity errors are blocking my development work on a new Pro 20x account.
Selected model is at capacity. Please try a different model.
@OpenAI_Support @OpenAIAPIhelper We can’t take it anymore; you can no longer get any work done at night.
For God’s sake, fix this error! You keep making ChatGPT worse with all these bugs and glitches, and there’s no one to respond or help resolve the issue—it’s absolutely outrageous!
Is there anyone in charge who can explain what is actually going on? Just look at how many users are affected by this bug and unable to get their work done.
Hi Selena (@OpenAI_Support),
I saw your offer to help review existing support cases involving persistent Codex capacity errors.
Could you please look into case #17875881? It was escalated for internal technical review on October 23, but I haven’t received any substantive update despite multiple follow-ups.
Support confirmed that my failed requests passed usage-limit checks before returning server-overload errors. The issue persists across multiple models, clients, and devices, including fresh sessions.
The same issue was also confirmed on my second account (case #18825931). A subsequent escalation attempt (case #18901118) was automatically closed.
I’ve documented the full technical evidence in the OpenAI Codex GitHub repository, under issue #53569.
Could you help confirm whether the engineering investigation is still active and arrange a review of potential account-level restrictions or routing issues?
Thank you for your help.
Hi @OpenAI_Support (Selena),
Could you please take a look at my existing support case #19097502?
I am on ChatGPT Pro 100, and ChatGPT Work has been consistently unusable since October 20, 2026. Every model returns “Selected model is at capacity. Please try a different model,” while regular ChatGPT works normally.
I have tried switching models and troubleshooting my network configuration without success.
I have received three AI-generated email responses, but the issue remains unresolved. The Help Center chat now also displays “You’ve reached the limit for new support conversations.”
Could you please help route my existing case for a human technical investigation? I can provide additional diagnostics privately.
Thank you.
mesmo problema @OpenAI_Support impossível gente pelo amor de Deus
Hi everyone! Thanks for sharing the details and your case IDs.
There was a known service issue affecting ChatGPT Work and Codex Web that has since been resolved.
If you’re still running into this error, try refreshing ChatGPT or starting a fresh session to see if that helps. You can also check OpenAI Status for any ongoing service issues.
If the error continues, please add any recent request IDs and failure timestamps, including your timezone, to your existing private support request. Let me know here too so I can follow up on your cases and look into the next steps.
-Selena
Hi @OpenAI_Support — I’m seeing the same issue reported here, and as of November 12, 2026, it is still unresolved for me.
I’m a ChatGPT Pro (20x) subscriber ($200/month). My Work/Codex usage is still around 90% remaining, yet reasoning and task execution in Work have clearly degraded, while regular Chat models in the same account and the same ChatGPT app work normally. I see the same degradation when I sign in to Codex with the same account.
Separately, using Amp Code’s official “Sign in with ChatGPT” subscription-sharing feature, sign-in succeeds but calling GPT-6.1 Sol returns HTTP 403 with error code subscription_sharing_user_not_eligible — despite the Pro 20x subscription being active and quota remaining.
I tried contacting OpenAI Support, but the Help Center says: “You’ve reached the limit for new support conversations. Please try again later.” So I currently cannot even open a new support conversation.
Could you please investigate: (1) why regular Chat works while Work/Codex degrade on the same account; (2) the model routing, reasoning configuration, subscription-sharing eligibility, server-side limits, and any possible account misconfiguration; (3) why an active Pro 20x account with ~90% quota remaining gets a subscription-sharing 403; and (4) if there is an account-side misconfiguration or mistaken restriction, please fix it. I also need a working route to human or technical support, since new support conversations are blocked.
I’m happy to share error screenshots, request timestamps, and conversation IDs privately. This does not look like quota exhaustion or a client-side issue — please help verify the actual cause. Thank you.
Quick follow-up on my case #19331381: another working day has passed with no reply or update from Support. The last message from Harry (Nov 11) said the technical review was still open and the refund was denied, but nothing since — and I still can’t open a new support conversation because of the limit. Work and Codex are still clearly degraded while regular Chat in the same account works fine, with about 90% of my Codex allowance still unused. Could someone from Support please give an actual update on the technical review?

