[App/iOS] Max reasoning effort missing for GPT-5.6 Luna, Terra, and Sol on macOS

Resolved 💬 2 comments Opened Jul 17, 2026 by conptz Closed Jul 17, 2026

Environment

  • Codex macOS app: 26.715.21425
  • Subscription: ChatGPT Pro
  • Mobile client: ChatGPT iOS app

Issue

The macOS model picker is missing the Max reasoning effort for all three GPT-5.6 models, even though Max is available for those models in the iOS app.

| Model | macOS behavior | iOS behavior |
|---|---|---|
| GPT-5.6 Luna | Stops at Extra High; Max is absent | Max is available |
| GPT-5.6 Terra | Goes directly from Extra High to Ultra; Max is absent | Max is available |
| GPT-5.6 Sol | Goes directly from Extra High to Ultra; Max is absent | Max is available |

Max should appear as a distinct reasoning level above Extra High and below Ultra. On macOS, Terra and Sol skip directly from Extra High to Ultra, while Luna stops at Extra High.

There is also a cross-client display mismatch: when Luna + Max is selected through the iPhone app, the same session appears as Luna Medium when viewed in the macOS app.

Steps to reproduce

  1. Sign in with a ChatGPT Pro account in Codex for macOS.
  2. Open the manual model and reasoning-effort dropdown.
  3. Select GPT-5.6 Luna and observe that the highest available effort is Extra High.
  4. Select GPT-5.6 Terra and observe that the options go from Extra High directly to Ultra.
  5. Select GPT-5.6 Sol and observe the same jump from Extra High directly to Ultra.
  6. Open the ChatGPT iOS app using the same account.
  7. Observe that Max is available for Luna, Terra, and Sol.

Expected behavior

  • Max should be selectable on macOS for Luna, Terra, and Sol, consistent with iOS.
  • The reasoning-effort order should be Extra High → Max → Ultra where Ultra is supported.
  • Selections should be preserved and displayed consistently across macOS and iOS.
  • The app should not silently display an iOS-selected Luna Max session as Luna Medium.

View original on GitHub ↗

This issue has 2 comments on GitHub. Read the full discussion on GitHub ↗