[Windows app] GPT-6 Astra reasoning ceiling differs across devices/new chats despite all efforts enabled #52860

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

26.1007.21434 (release date 2026/10/09)

What subscription do you have?

ChatGPT Plus

What platform is your computer?

Windows 11 x64

What issue are you seeing?

The Windows Codex/ChatGPT desktop app exposes inconsistent reasoning-effort ceilings for the same account and app version.

On one Windows PC, a brand-new Codex chat with GPT-6 Astra shows Medium as the rightmost / maximum selectable reasoning effort, even though:

• Settings → Configuration → Model features → Available reasoning efforts has all six levels enabled: Low, Medium, High, Extra High, Max, Ultra.
• In the same fresh-chat UI, GPT-6.1 Sol exposes additional reasoning positions beyond Medium.
• On another Windows PC signed into the same ChatGPT account and running the same app version (26.1007.21434), GPT-6 Astra can be selected at Ultra.
• Existing chats on the same machine can also retain/display different model + reasoning combinations (e.g. Astra Medium, Astra High, Astra Ultra, Sol Ultra), so the behavior is not explained solely by the global Available reasoning efforts setting.

This appears to be a Windows desktop model-catalog / reasoning-metadata synchronization or UI-filtering issue rather than a hardware capability difference.

The account was at a Codex/Work usage limit during some observations, but that does not fully explain the behavior: in the same usage-limit state, GPT-6.1 Sol still displayed higher reasoning positions while GPT-6 Astra was capped at Medium.

Steps to reproduce

1. Sign into ChatGPT Plus in the Windows Codex/ChatGPT desktop app.
2. Open Settings → Configuration → Model features.
3. Enable all six Available reasoning efforts (Low, Medium, High, Extra High, Max, Ultra).
4. Create a brand-new Codex chat.
5. Select GPT-6 Astra.
6. Open the reasoning-effort slider/dropdown.
7. Observe that Medium may appear as the maximum/rightmost selectable effort.
8. Switch the same new chat to GPT-6.1 Sol.
9. Observe that additional reasoning positions are available beyond Medium.
10. Compare with another Windows device on the same account/app version where GPT-6 Astra Ultra is available.

Expected behavior

If GPT-6 Astra supports higher reasoning efforts for this account, the Windows app should expose the same available reasoning levels consistently across devices and new chats, subject only to documented plan/model restrictions.

If a server-side rollout, quota state, or model-specific restriction intentionally limits Astra to Medium, the UI should make that explicit instead of presenting a device/session-dependent ceiling that conflicts with the enabled “Available reasoning efforts” setting.

Additional information

• Same ChatGPT account across devices.
• Same desktop build reproduced with different Astra ceilings.
• App reports itself up to date.
• Screenshots are available showing:
  1. all six reasoning efforts enabled in Settings;
  2. new GPT-6 Astra chat capped at Medium;
  3. same new-chat screen with GPT-6.1 Sol exposing higher effort positions;
  4. another Windows PC on the same account/app version with GPT-6 Astra Ultra available.

Related reports that may be relevant:

• Android Remote reasoning-effort mismatch: mobile Max/Ultra appears as Medium on Windows Desktop #42301 — reasoning-effort mismatch between clients
• [App] Advanced power slider skips Max and mixes model, reasoning, and Ultra #31968 — reasoning slider / supported-level UI mismatch
• codex app does not respect model_reasoning_effort #10643 — app not respecting model_reasoning_effort (older/closed)