[BUG] Scheduled task model selection reverts to Sonnet on next automated run

Preflight Checklist

• [x] I have searched existing issues 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?

When editing a scheduled task in the Claude Code Desktop app (Schedule page), I changed the model from Sonnet 4.6 to Opus. The change appeared to save successfully.

However, on the next automated run (the following morning), the task ran using Sonnet 4.6 instead of Opus. The model selection did not persist between runs.

Steps to reproduce:

1. Open Claude Code Desktop → Schedule page
2. Click on an existing scheduled task (e.g., a daily recurring task)
3. Click Edit
4. Change the model selector from Sonnet to Opus
5. Save the task
6. Wait for the next automated run to trigger

Actual behavior: The task runs using Sonnet 4.6, ignoring the Opus selection made in step 4. The model selector in the active session shows "Sonnet 4.6" in the bottom-right corner.

This was reproduced across multiple scheduled tasks (tried setting all 4 of my scheduled tasks to Opus; all reverted to Sonnet on next run).

What Should Happen?

The scheduled task should run using Opus (the model selected and saved in the task configuration). The model selection should persist across automated runs.

Error Messages/Logs

No error messages. The task runs successfully but on the wrong model (Sonnet 4.6 instead of Opus).

Environment:
- Claude Code Desktop on Windows 11 Pro (10.0.26200)
- Latest version as of April 7, 2026
- 4 scheduled tasks configured (daily briefing, post-meeting followup, weekly sweep, quarterly audit) — all exhibit the same behavior

Steps to Reproduce

1. Open Claude Code Desktop app
2. Go to Schedule page (left sidebar)
3. Click on an existing scheduled task
4. Click Edit
5. Change the model selector from Sonnet 4.6 to Opus
6. Save the task
7. Confirm the task shows the updated model
8. Wait for the next automated run to trigger (e.g., next morning for a daily task)
9. Observe that the task session shows "Sonnet 4.6" in the bottom-right corner, not Opus

Claude Model

Opus

Is this a regression?

No, this never worked

Last Working Version

_No response_

Claude Code Version

Claude Code Desktop (latest as of April 7, 2026)

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

PowerShell

Additional Information

This bug occurs on the Claude Code Desktop app (not CLI). The scheduled tasks are configured via the Desktop Schedule page. The model reverts on every automated run — manually switching the model in an active session works fine, but the saved preference doesn't carry over to the next scheduled execution.

Note: This is specifically about the Desktop scheduled tasks feature, not the CLI /schedule command.