Skip to content

Desktop app does not remember selected thinking effort across new chats like CLI does #20191

Closed as not planned

Description

@miaopan607

Description

CLI remembers the selected thinking effort for a model across new chats, but the Desktop app does not.

If I select xhigh for a reasoning model in CLI, the next new chat still shows xhigh. In the Desktop app, selecting the same thinking effort and then starting a new chat resets the control back to Default.

This makes Desktop inconsistent with CLI and makes it hard to trust what the Desktop UI is showing.

This is different from #17469, which was about switching agents resetting the thinking level.

Plugins

None

OpenCode version

1.3.9

Steps to reproduce

  1. In CLI, pick a model with thinking variants and select xhigh
  2. Start another new chat/session in CLI and observe that xhigh is still selected
  3. In the Desktop app, pick the same model and select xhigh
  4. Start a new chat in the Desktop app
  5. Observe that the thinking effort resets to Default

Screenshot and/or share link

No share link. Reproduced locally.

Operating System

Windows 10

Terminal

CLI: Windows Terminal
Desktop: N/A

Activity

  1. github-actions commented on Mar 31, 2026

    @github-actions
    Contributor

    This issue might be a duplicate of existing issues. Please check:

    Since you already referenced #17588, it may be worth confirming with maintainers whether this desktop-specific behavior warrants a separate issue or should be tracked under #17588.

  2. changed the title [-]Desktop app does not reflect configured reasoning effort from model options on new chat[/-] [+]Desktop app does not remember selected thinking effort across new chats like CLI does[/+] on Mar 31, 2026
  3. brandon-julio-t commented on Apr 16, 2026

    @brandon-julio-t
    Contributor

    Same issue here. I have submitted a PR to fix this, hope it helps, would love to get this fixed to maintain my flow state.

    Adding more details from my side

    Description

    In web and desktop, if I pick a non-default reasoning effort for a model and then start a new session, the effort resets back to Default.

    TUI persists the selected effort across sessions, so web/desktop behavior is inconsistent.

    Expected:

    • New web/desktop sessions keep the last selected reasoning effort for the current model, like TUI does.

    Actual:

    • New web/desktop sessions reset the reasoning effort to Default.

    Plugins

    None

    OpenCode version

    1.4.6

    Steps to reproduce

    1. Open OpenCode web or desktop.
    2. Select a model that supports reasoning effort / variants.
    3. Change the reasoning effort to a non-default value such as high.
    4. Create a new session.
    5. Observe that the reasoning effort resets to Default.

    Screenshot and/or share link

    No response

    Operating System

    macOS 26.4.1

    Terminal

    Ghostty

    Compare TUI vs OpenCode Web/Desktop:

    • In TUI, after visiting a thread with xhigh model variant, creating new thread will re-use the xhigh variant
    tui.mp4
    • In web/desktop, after visiting a thread with xhigh model variant, creating new thread does not re-use the xhigh variant
    before.mp4
  4. removed
    webRelates to opencode on web / desktop
    on May 3, 2026
  5. github-actions commented on Jul 4, 2026

    @github-actions
    Contributor

    To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions