Service eligibility and regional restrictions
PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.
Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.
Key Takeaways
- Tibo labels his October 15 post Day 11: a series of changes and additional compute is intended to significantly reduce the model-capacity message.
- If it still happens, he says the task should eventually continue by itself rather than remain blocked. That is not a promise of immediate recovery.
- The post does not specify a recovery window, affected models or accounts, client, region, or rollout schedule.
- It makes no claim about pricing, usage, or quota resets.
What did the Day 11 update say?
At 15:11 UTC on October 15, 2026, Tibo (@thsottiaux) posted an update he labels Day 11. He says a series of changes has landed and more compute has been added to significantly reduce the disruptive message, “Selected model is at capacity. Please try a different model.”
He adds that if the message still appears, the task should eventually continue by itself instead of stopping at the error. Tibo also asks people who still see the issue to let him know.
This is a mitigation update about a specific model-capacity message, not an announcement that the issue has been eliminated. The post describes the intended improvement but gives no before-and-after error rate or service guarantee.
If the message appears, check the original task first
If you see that exact message, return to the original conversation and check its task status. Because Tibo describes automatic continuation as the goal, a cautious next step is to watch the original task before starting an identical one that could duplicate work. This is practical advice, not a new retry rule published in the post.
Hypothetical example: a task briefly shows the capacity message, then the same conversation starts producing output again. That would fit the stated direction of automatic resumption. This is an illustrative scenario, not a test, and it does not show that every error will recover this way.
If the task remains stopped, you can report the exact message, approximate time, and the entry point or model you can identify through Tibo's post. Do not publish API keys, proxy credentials, or a full conversation containing private information.
What timing and coverage remain unknown?
The post does not say how long automatic continuation usually takes, whether a user must click anything, or which task state is preserved. It lists no affected models, plans, countries, or clients, and gives no staged rollout schedule.
The supported summary is narrow: Tibo says changes have landed and more compute has been added to reduce this particular message and let affected tasks eventually continue. That does not establish that every Codex error is fixed, that every account has the change immediately, or that the message will never return.
The post says nothing about prices, task usage, or quota resets. Automatic continuation does not imply that a task is free, used quota is restored, or plan entitlements have changed.
Is the issue fixed, or should it happen less often?
The post is best read as “reduce how often it occurs and improve what happens when it does.” Tibo says the changes should significantly reduce the message and that a task should eventually resume; he does not promise to remove it entirely or guarantee every recovery.
If the same message appears in your session, first check whether the original task continues. If it stays stopped, report it through the post's invitation. Since no standard timeout or recovery button is specified, do not present a self-chosen number of waiting minutes as official guidance.
Sources
Frequently Asked Questions
Has the Codex “Selected model is at capacity” error been completely fixed?
The post says changes and additional compute should reduce this message and let an affected task eventually continue. It does not promise to eliminate the message or provide error-rate or recovery-success data.
Should I immediately submit the task again when I see it?
The post does not prescribe a retry procedure. Checking the original conversation first is a cautious way to avoid duplicate work; if the task remains stopped, you can report it through Tibo's post.
Which plans, models, or countries does this change cover?
The post does not identify plans, models, clients, or regions, and it gives no rollout schedule. It cannot yet be summarized as active for every account.