Skip to content

Significant response time degradation over last releases #2684

Closed

Description

@fkukuck

I'm experiencing a significant performance degradation of how fast opencode works in the last versions.
These recordings should give you an impression.

v0.5.24.mp4
v0.10.1_v3.mov

They are unaltered, and reproducible every single time under these conditions (haven't tried others):

  • gpt-5-mini
  • GitHub Copilot Authentication
  • v0.5.24 vs. v0.10.1

Activity

  1. changed the title [-]Significant performance degradation over last releases[/-] [+]Significant response time degradation over last releases[/+] on Sep 19, 2025
  2. rekram1-node commented on Sep 19, 2025

    @rekram1-node
    Contributor

    @fkukuck I think this is just the default settings giving to gpt-5 models changing, right now it defaults to high reasoning effort which means it takes longer

    You can override this if you would like

  3. self-assigned this
    on Sep 19, 2025
  4. fkukuck commented on Sep 19, 2025

    @fkukuck
    ContributorAuthor

    I wasn’t aware since I thought GitHub Copilot always runs medium and there is no choice possible.

  5. fkukuck commented on Sep 19, 2025

    @fkukuck
    ContributorAuthor

    Will confirm next week

  6. rekram1-node commented on Sep 19, 2025

    @rekram1-node
    Contributor

    @fkukuck to override this you can set this in your opencode.json:

    {
      "$schema": "https://opencode.ai/config.json",
      "provider": {
        "github-copilot": {
          "models": {
            "gpt-5-mini": {
              "options": {
                "reasoningEffort": "minimal",
                "textVerbosity": "low"
              }
            }
          }
        }
      }
    }
    

    to see available options read vercel ai sdk docs: https://ai-sdk.dev/providers/ai-sdk-providers/openai#responses-models

    reasoningEffort 'minimal' | 'low' | 'medium' | 'high' Reasoning effort for reasoning models. Defaults to medium.
    textVerbosity 'low' | 'medium' | 'high' Controls the verbosity of the model's response. Lower values result in more concise responses, while higher values result in more verbose responses. Defaults to 'medium'.

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