GPT-5.6 Sol and GPT-6 Astra rejected in Codex on active ChatGPT Pro account

I have an active ChatGPT Pro subscription, but Codex rejects GPT-5.6 Sol and GPT-6 Astra with:

"The 'gpt-5.6-sol' model is not supported when using Codex with a ChatGPT account."

and

"The 'gpt-6-astra' model is not supported when using Codex with a ChatGPT account."

GPT-5.6 Terra works normally on the same account.

ChatGPT Desktop Settings → Usage shows:

• Current plan: Pro
• Weekly allowance: 80% remaining

ChatGPT Work is available on chatgpt.com for the same Personal account.

I have already:

• Updated ChatGPT Desktop
• Signed out and reauthenticated ChatGPT/Codex
• Explicitly selected my Personal workspace
• Fully restarted the desktop app
• Removed custom model overrides
• Confirmed Terra works normally

OpenAI Support indicated this appears to be a backend access/gating issue.

Diagnostic information

• Session ID: 01a0b135-5358-7683-a2d1-338a1f785e04
• Failure time: 2026-09-18 05:06 UTC+8
• ChatGPT Desktop version: 26.911.61220
• OS: Windows
• Active workspace: Personal
• ChatGPT plan: Pro
• Weekly allowance remaining: 80%

Affected models

• gpt-5.6-sol — rejected
• gpt-6-astra — rejected
• gpt-5.6-terra — works normally

Steps to reproduce

1. Sign in to ChatGPT Desktop with my active ChatGPT Pro account.
2. Switch to Codex in the Personal workspace.
3. Select GPT-5.6 Sol.
4. Start a new conversation and send "hello".
5. Codex returns:
   "The 'gpt-5.6-sol' model is not supported when using Codex with a ChatGPT account."
6. Switch to GPT-6 Astra and repeat.
7. Astra returns the same type of "not supported" error.
8. Switch to GPT-5.6 Terra.
9. Terra works normally.

Expected behavior

GPT-5.6 Sol and GPT-6 Astra should be available in Codex for my active ChatGPT Pro account.

No repository file or test is identified. Start by reproducing the failure in Codex through ChatGPT Desktop with the Personal workspace and compare Sol and Astra against the working Terra model, using the listed version, OS, session ID, and account diagnostics. Done means the entitlement behavior is corrected or the supported-model policy is clearly documented.