GPT-5.6 Sol repeatedly hits “Selected model is at capacity” in Codex Desktop

Codex
21

The same thing is happening to me, and now I’m stuck with Terra. It’s not bad, but I have to manually verify every step, which is exhausting.

22

Thank you to everyone who has reported experiencing the same problem.

Note: I am not an OpenAI employee, just a user of the forum.

OpenAI is working to investigate and resolve the issue, but they need additional information from those affected. Please include the following details in your reply:

23

Additionally, could you share the Feedback ID or Request ID with me? That would really help with the investigation.

Avinash

24

019fa9cd-775c-73f3-a23b-982d922481ae

If there isn’t enough capacity to support the service reliably, perhaps it would be worth reconsidering whether it should continue being sold as if it were fully available.

I have a total allowance of 100 million tokens, yet my task has been running for two days and has been interrupted repeatedly by the same capacity error. At this point, I’m genuinely curious how users are expected to complete any serious work under these conditions.

I paid for a membership to use the service, not to spend my time repeatedly looking at error messages. A paid subscription should provide access to a functioning product, not simply a more consistent view of the same failure screen.

The issue has now persisted for two full days. One would assume that a capacity problem affecting paid users this consistently would receive rather urgent attention, but apparently that expectation may have been overly optimistic.

Could you please explain when this issue will actually be resolved, rather than simply asking users to keep trying again?

25

I’m seeing similar reports from other users, so it may be a temporary capacity issue rather than something specific to your setup. Since you’ve already submitted feedback and contacted support, you’ve covered the right steps. Hopefully OpenAI can confirm whether there’s still an ongoing backend capacity issue with GPT-5.6 Sol. In the meantime, trying another model is probably the best workaround until the service stabilizes.