GPT-5.6 Sol capacity errors followed by auth.openai.com HTTP 500 when switching accounts

What issue are you seeing?

On August 31, 2026, GPT-5.6 Sol in Codex repeatedly started failing with:

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

This appears similar to #41790 and other reports from today.

After the usage/credits on one paid account were exhausted, I tried to switch Codex to a second paid OpenAI account. The sign-in flow then became unusable: auth.openai.com/choose-an-account repeatedly returns a Server error 500.

The 500 error is not limited to one account. I reproduced it while trying to sign in with multiple OpenAI accounts, so this does not look like an account-specific subscription or credential problem.

At the moment this creates two separate blockers in the same workflow:

1. GPT-5.6 Sol is frequently unavailable because of model-capacity errors.
2. Switching to another paid account is impossible because OpenAI Auth fails with HTTP 500 during account selection.

Steps to reproduce

1. Use Codex with GPT-5.6 Sol.
2. Encounter repeated Selected model is at capacity. Please try a different model. errors.
3. Sign out / attempt to switch to another paid OpenAI account.
4. Proceed to the OpenAI account-selection page.
5. auth.openai.com/choose-an-account returns Server error 500.
6. Retry with a different OpenAI account.
7. The same 500 error occurs again.

Expected behavior

• Temporary GPT-5.6 Sol capacity pressure should recover or be clearly reported as a service-side capacity incident.
• The OpenAI authentication flow should still allow switching to another valid paid account.
• An account-selection request should not fail with HTTP 500.

Actual behavior

• GPT-5.6 Sol repeatedly reports that the selected model is at capacity.
• The account-switch/login flow fails server-side at auth.openai.com/choose-an-account with Server error 500.
• The 500 is reproducible with multiple accounts.

Additional information

• Date: August 31, 2026
• A screenshot of the OpenAI Auth Server error 500 page is available if maintainers need it.
• This is distinct from a wrong-password or single-account issue because the same server error occurs with different accounts.
• Possibly related capacity reports: #41790, #41805, #41810.

Please investigate whether the authentication failure is related to the current Codex/model-capacity degradation or is a separate OpenAI Auth incident.

The report names no repository file, test, or code entry point. Start by separating the GPT-5.6 Sol capacity reports from the auth.openai.com/choose-an-account HTTP 500, then determine whether either failure is actionable in Codex or is an external service incident; done means the scope and ownership of the problem are established.