Skip to content

[FEATURE]:Show reasoning/variant level for subagents in the UI #26266

Closed as not planned

Description

@zzy2210

Feature hasn't been suggested before.

  • I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

Hi, I’d like to ask whether subagents could display their reasoning level.

Currently, when a subagent is invoked, only the model name is shown, but the reasoning level is not displayed. This can make it feel a bit uncertain or less reassuring, since it’s hard to tell what reasoning setting the subagent is using.

Would it be possible to add support for showing the reasoning level when a subagent is called?

Thanks!

Activity

  1. github-actions commented on May 8, 2026

    @github-actions
    Contributor

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

  2. zzy2210 commented on May 8, 2026

    @zzy2210
    Author

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

    * [[FEATURE]: Show subagent model in session tree / subagent view #23028](https://github.com/anomalyco/opencode/issues/23028): Requests showing the subagent's model in the session tree/UI — your request to also show the reasoning/variant level is closely related and may be better added as a comment there.
    
    * [Subtasks do not inherit the active model variant from the parent session #24757](https://github.com/anomalyco/opencode/issues/24757): Subtasks not inheriting the active model variant from the parent session — related to variant/reasoning level handling in subagents.
    

    I think this feature would be even more useful if the UI showed not only the subagent's model, but also the effective reasoning / variant / effort level being used.

    For example:

    • Plan — GPT-5.4 — high
      • General — GPT-5.4-mini — medium
      • Build — Sonnet-4.5 — default

    Currently, when a subagent is invoked, seeing only the model name can still feel incomplete, especially for models where variants or reasoning-effort levels significantly affect behavior, latency, and cost. Displaying the effective variant/reasoning level would make subagent runs more transparent and easier to audit/debug.

    This also seems related to #24757, where subtasks may not inherit the active model variant from the parent session. Even after inheritance/configuration behavior is fixed, showing the effective variant in the session tree or subagent view would help users verify what is actually being used at runtime.

  3. mewa commented on Jun 26, 2026

    @mewa

    Agree. It's really unintuitive that the subagent is missing the model variant. It should be showing the exact same way the top-level UI already does.

  4. akhansari commented on Jul 15, 2026

    @akhansari

    This feature will be really helpful. Especially because it's not the same cost between medium and xhigh variants. It's reassuring to verify that and avoid accidentally burning tokens.

  5. samiralibabic commented on Jul 23, 2026

    @samiralibabic

    I can reproduce this in the current TUI.

    The main-session prompt footer shows the selected model variant, but when navigating into a subagent session, the TUI replaces that footer with SubagentFooter. That footer currently shows only the subagent name, position, context usage, cost, and navigation controls—no model or variant.

    It would be useful to display something like:

    Explore · gpt-5.4 · xhigh

    The value should be derived from the child session’s effective persisted model/variant, rather than the parent/global model selector, so it remains accurate when the subagent has its own model or variant configuration.

    This appears to be a presentation-only gap: the child session already records the effective model/variant.

  6. github-actions commented on Sep 22, 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.

  7. samiralibabic commented on Sep 29, 2026

    @samiralibabic

    Why "open a new one" and not "re-open it"?! Stupid bot!

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