Selected model is at capacity. Please try a different model.

English translation:

What version of the Codex App are you using (From “About Codex” dialog)?

26.803.81509

What subscription do you have?

Why does this error keep appearing?

What platform is your computer?

_No response_

What issue are you seeing?

Why does this error keep appearing?

What steps can reproduce the bug?

Why does this error keep appearing?

What is the expected behavior?

_No response_

Additional information

_No response_

<!-- codex-issue-translator -->

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

• #37935
• #37795
• #37790

Powered by Codex Action

为什么总出现这个问题？ +1

same

Additional reproducible Windows report specifically for GPT-5.6 Sol + Ultra. Switching to Terra/Luna is not an acceptable workaround for this workload; the user needs the selected Sol + Ultra configuration to work.

Environment

• Codex App (Microsoft Store package): 26.803.10989.0
• Bundled Codex CLI: 0.147.0-alpha.6.6
• Platform: Microsoft Windows NT 10.0.26200.0 x64
• Active user configuration:
• model = gpt-5.6-sol
• model_reasoning_effort = ultra
• service_tier = priority

Observed behavior

• Exact error: Selected model is at capacity. Please try a different model.
• The error occurs on ordinary prompts and across newly started tasks; it is not tied to one prompt or repository.
• A sanitized read-only query of the local Codex diagnostic database found 44 turn-level Sol + Ultra capacity-error events across 35 task IDs between 2026-08-12 15:09:55 and 17:02:17 Asia/Taipei (UTC+8).
• The app remained authenticated and continued refreshing its model catalog, so this does not look like a general sign-in or offline failure.

Expected behavior
A supported gpt-5.6-sol + ultra request should execute successfully. If the model is temporarily saturated, Codex should preserve the turn and transparently queue/retry with bounded backoff instead of terminating the task or requiring a model downgrade. The UI should also distinguish model capacity from account quota/authentication errors.

Please investigate the Sol + Ultra capacity/routing path and desktop-client retry behavior. Raw logs were not posted publicly because they may contain sensitive session content; they can be shared through the private /feedback flow if requested by a maintainer.

<!-- codex-sol-ultra-capacity-report-2026-08-12 -->