I strongly suspect that OpenAI has imposed these restrictions out of risk prevention. It aims to avoid the scenario where newly subscribed Pro accounts exhaust usage quotas quickly and then apply for refunds. However, such blanket restrictions are extremely unfair to legitimate regular users and greatly hinder our normal use.
Is anyone else experiencing serious intermittent capacity issues with GPT-5.6 Sol on Codex, especially on Pro 20x?
I ran several tests on the same Pro 20x account:
• Sol / xhigh + Fast mode → worked for 30+ minutes, then eventually failed ❌
• Sol / High → failed after 4m35s ❌
• Sol / xhigh → failed immediately after submission ❌
All three eventually ended with:
“Selected model is at capacity. Please try a different model.”
I’ve also seen:
Error running remote compact task: Selected model is at capacity
and local logs show:
codex_error_info: server_overloaded
The behavior is highly intermittent. ChatGPT Web with GPT-5.6 Sol also failed with a capacity error, then worked again later in a new conversation.
Even stranger: a Plus account on the same PC/network was able to use Sol normally, while my Pro 20x account repeatedly hit these errors.
OpenAI Support has already confirmed that my Codex usage state does not appear to be limit-blocked.
Case #12953300.
This looks much more like an intermittent backend capacity / routing / scheduling issue than a normal usage limit.
Is anyone else seeing the same behavior, especially Pro users?
@OpenAI@OpenAIDevs — could the Codex team please investigate?
I think this explanation is totally unreasonable. I logged in with my Plus account on the exact same device and ran it for a long time without running into this issue at all. But when I switched to my Pro account, it stopped working properly. Are Pro users supposed to have worse access privileges than Plus users? This problem has been troubling me for ages, and I still cannot get my work done normally to this day.
Update: the Sol xhigh + Fast task that initially survived for 30+ minutes has now failed with the same capacity error too.
So all 3 Codex Sol tasks eventually failed — just at very different times: immediately, after ~4m35s, and after 30+ min.
@kshern888@thsottiaux Same issue here. Pro 20x repeatedly gets “Selected model is at capacity”, while Plus works fine on the same PC/network with GPT-5.6 Sol. Often happens during remote compact.
@thsottiaux Hi @thsottiaux — Codex GPT-5.6 Sol on my Pro 20x account keeps showing “Selected model is at capacity,” often during remote compact. Same PC/network with a Plus account works fine. Possible account/tier routing issue?
Hi @thsottiaux@OpenAIDevs — Codex GPT-5.6 Sol on my Pro 20x account keeps showing “Selected model is at capacity,” often during remote compact. Same PC/network with a Plus account works fine. Possible account/tier routing issue?