[BUG] Claude Code is showing one model selected while running another

### Preflight Checklist

- [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code

### What's Wrong?

Claude Code is showing one model selected while running another.

Desktop app shows: Fable 5.1 (extra effort), selected ~2+ hours ago. The session reported to the model: Opus 5, continuously, for ~8 hours.
Switch point: a model change landed mid-session at the moment I was issuing prompts from the iPhone client earlier in the day. After returning to the desktop and selecting Fable 5.1, the session kept reporting Opus 5 in spite of showing Fable 5.1 the entire time.

Evidence: git commit trailers in the repo are harness-generated and flip from "Co-Authored-By: Claude Fable 5.1" to
"Co-Authored-By: Claude Opus 5" at 12:50 and never flip back, including on work done from the desktop hours later.

Question: does a model selected on one device propagate to a session already set to a different model on another device?

### What Should Happen?

The model shown as selected in the Claude Code client should be the model that executes the session's prompts, including after the session's model was changed from another device. The displayed model and the executing model should never diverge silently.

### Error Messages/Logs

```shell

```

### Steps to Reproduce

Using the /remote feature on the Claude iOS app, then switching to Claude on Windows.

### Claude Model

None

### Is this a regression?

Yes, this worked in a previous version

### Last Working Version

_No response_

### Claude Code Version

Claude desktop app 2.110.0

### Platform

Anthropic API

### Operating System

Windows

### Terminal/Shell

PowerShell

### Additional Information

_No response_

Start by reproducing the /remote flow from the Claude iOS app, then switch to Claude on Windows and compare the displayed model with the model reported by the session and commit trailers. Done means a model change from another device cannot leave the selected and executing models silently diverged.

Written by the indexing model from the issue text.