iOS Remote forces official model selection when continuing a custom-model task

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

ChatGPT Desktop 26.810.52044 (build 6662); ChatGPT iOS 1.2026.209 (30960470514)

What subscription do you have?

Enterprise

What platform is your computer?

Host: macOS 26.5.2 (25F84), Apple Silicon (Darwin 25.5.0 arm64). Client: ChatGPT for iOS.

What issue are you seeing?

When continuing an existing Codex task from ChatGPT iOS Remote, the mobile app requires selecting a model before the next message can be sent.

The existing task is running on the connected Mac with a custom model named codex/gpt-5.6-sol. This is a user-defined model name backed by a custom model provider; it is distinct from OpenAI's official gpt-5.6-sol model.

The iOS model picker only lists official models, including 5.6 Sol. It does not list the connected host's custom model codex/gpt-5.6-sol, and it does not offer a "keep current model" option. Because model selection is mandatory, choosing 5.6 Sol changes the next turn to the official gpt-5.6-sol model instead of preserving the task's custom model/provider.

As a result, a task that works with the custom model on Desktop cannot be continued from iOS without silently changing its model/provider.

What steps can reproduce the bug?

1. On the Mac host, configure a custom model provider and a custom model named codex/gpt-5.6-sol.
2. In ChatGPT Desktop, create a Codex task using codex/gpt-5.6-sol and verify that the task runs with that custom model.
3. Connect ChatGPT for iOS to the Mac through Remote.
4. Open the existing task on iOS and try to send a follow-up message.
5. Observe that iOS requires a model selection.
6. Open the model picker. It lists the official 5.6 Sol model, but not the host's custom codex/gpt-5.6-sol model and no option to preserve the current selection.
7. Select 5.6 Sol and send the follow-up.
8. Observe that the task now uses the official gpt-5.6-sol model rather than the original custom model/provider.

This is reproducible when continuing an existing task; it is not limited to creating a new task.

What is the expected behavior?

Continuing an existing Remote task should preserve the task's current model and provider by default.

At least one of the following should be supported:

• Do not require model selection when continuing an existing task; keep its persisted model/provider.
• Include the connected host's custom model catalog in the iOS picker, so codex/gpt-5.6-sol can be selected explicitly.
• Provide a "keep current model" option.

The iOS client should not replace a custom model/provider with an official model merely because the names are similar.

Additional information

The model names are intentionally similar but refer to different targets:

• codex/gpt-5.6-sol: user-defined custom model name on the connected host.
• gpt-5.6-sol / "5.6 Sol": official model shown by the iOS picker.

No custom provider URL, credentials, organization details, or private task content are included.

Related but distinct: #24780 concerns custom-provider tasks disappearing from the mobile thread list. This report concerns mandatory model selection while continuing a visible existing task and the resulting model/provider override.