Skip to content

[FEATURE]: show tokens / second #5374

Closed

Description

@IceWreck

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

Display current and average tokens/s for requests - will be ideal for comparing performance across different providers and models.

Activity

  1. added
    discussionUsed for feature requests, proposals, ideas, etc. Open discussion
    on Dec 11, 2025
  2. github-actions commented on Dec 11, 2025

    @github-actions
    Contributor

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

    These issues are related to displaying performance metrics and detailed session information. Tokens/second metrics could be part of an expanded performance monitoring feature alongside these existing requests.

    Feel free to ignore if your specific case differs from these.

  3. Madmanden commented on Dec 25, 2025

    @Madmanden

    This is a feature I'd love. I'm testing different models, and speed is one of the factors I'm comparing. A token speed indicator would be very helpful.

  4. ForbiddenEra commented on Jan 23, 2026

    @ForbiddenEra

    plz+1

  5. haydonryan commented on Feb 5, 2026

    @haydonryan

    Huge plus one from me. Just approx t/s in the bottom right corner or something.

    It would really help people who are looking to buy hardware to run models locally to know what experience they're seeing in the cloud. This can then be cross referenced to other's builds.

  6. JohnCode-Dev commented on Feb 10, 2026

    @JohnCode-Dev

    Hey all — I just opened a PR that tackles this: #12721

    It adds a tok/s metric right after the duration in the response footer (e.g. 18.3s · 131 tok/s). I've been switching between providers a lot and wanted a quick way to see the actual speed difference without external tools.

    The TPS logic lives in core/tokens/ so it can be reused outside the TUI later. Would love feedback from anyone still interested in this!

  7. SamWang32191 commented on Mar 7, 2026

    @SamWang32191

    This feature would be very helpful for my use case. I was wondering if there have been any updates on it.🙏

  8. ForbiddenEra commented on Mar 10, 2026

    @ForbiddenEra

    This would be great for people doing local stuff

  9. Navaneeth-Sharma commented on Mar 13, 2026

    @Navaneeth-Sharma

    Yes, adding to the above conv. This feature would be great.

  10. evilbuck commented on Mar 19, 2026

    @evilbuck

    +1 on this. If nothing else, just to keep the providers honest.

  11. slimypi commented on Mar 26, 2026

    @slimypi

    +1 This would be great to see if providers are not lying...

  12. ForbiddenEra commented on Mar 27, 2026

    @ForbiddenEra

    +1 This would be great to see if providers are not lying...

    I suppose that's a good use too!

  13. PEMessage commented on Mar 30, 2026

    @PEMessage

    +1 on this. very helpful

  14. codenoid commented on Mar 31, 2026

    @codenoid

    +1 on this

  15. Tarquinen commented on Apr 4, 2026

    @Tarquinen
    Contributor

    This can be done via plugin now, for example:

    https://github.com/Tarquinen/oc-tps

  16. 3 remaining items

  17. SergioSV96 commented on May 7, 2026

    @SergioSV96

    This should be part of native OpenCode! vouch

  18. chris-hatton commented on Jun 16, 2026

    @chris-hatton

    This should be built-in - surely everyone wants to know how their compute endpoint is performing during normal work.

  19. chinhnt38 commented on Jul 3, 2026

    @chinhnt38

    +1 vote for this

  20. bzavolovich commented on Jul 6, 2026

    @bzavolovich

    +1 on this

  21. grahammiranda13 commented on Aug 10, 2026

    @grahammiranda13

    +1 on this

  22. JasonJason-oy commented on Sep 8, 2026

    @JasonJason-oy

    https://github.com/JasonJason-oy/oc-metricboard
    try this, for real-time LLM metrics (TPS / TTFT / token usage) with sub-agent aggregation and per-model breakdown.

  23. yisier commented on Sep 11, 2026

    @yisier

    +1 on this

  24. JoyousJohn commented on Sep 22, 2026

    @JoyousJohn

    +1

  25. OverHash commented on Sep 22, 2026

    @OverHash

    It hasn't been mentioned in the thread, but this is a native feature of OpenCode V2 thanks to #45265!

    In cli.json(c):

    {
      "session": {
        "tps": true
      }
    }

    which displays tokens per second for each step.

  26. rekram1-node commented on Oct 1, 2026

    @rekram1-node
    Contributor

    https://opencode.ai/v2/docs

    as said by @OverHash this is supported in opencode v2 out of the box!

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