Skip to content

Qwen3 Coder #1251

Closed
Closed

Description

@JC1738

Looks like Qwen team forked Gemini-cli to offer better support for the new Qwen3 coder model. Any plans to pull in similar changes? I know OpenRouter is already supported, but I am guessing there is model handling that needs tweaked to optimize for Qwen3 output?

Activity

  1. self-assigned this
    on Jul 23, 2025
  2. bigcodegen commented on Jul 25, 2025

    @bigcodegen

    I guess this needs to be fit in somehow: https://huggingface.co/Qwen/Qwen3-Coder-480B-A35B-Instruct/blob/main/qwen3coder_tool_parser.py

    A bit unusual of a need.

    Edit: Nevermind, seems providers are taking care of this mostly

  3. ratacat commented on Jul 25, 2025

    @ratacat

    I've been having trouble getting qwen3 coder to run over lmstudio...it works fine, can query and curl and use other clients just fine, but when I try to run query from Opencode it just hangs, LM studio says it replied, and Opencode says 'working'...

  4. fungilation commented on Jul 25, 2025

    @fungilation

    also generally, temperature config per model/provider used? https://www.youtube.com/watch?v=gBuuaAX4ec8 notes he had to use Roocode instead of Opencode as he couldn't specify temp 0.7 which is optimal for Qwen3 Coder

  5. Syazvinski commented on Jul 31, 2025

    @Syazvinski

    Im having same issue with LM studio and tool calls. Anyone have any pointers? Using the new qwen coder a3b 30b, and 400b have the same issues where tool calls dont work.

  6. esafak commented on Jul 31, 2025

    @esafak
    Contributor

    The Z.AI GLM models are much more reliable.

  7. grigio commented on Aug 4, 2025

    @grigio

    I noticed that Qwen3 235B A22 works via Openrouter, but smaller models don't

  8. geoh commented on Oct 16, 2025

    @geoh

    I've just created an OAuth plugin that allows you to use the Qwen3 Coder model via the qwen.ai authentication which offers 2000 free reqeusts per day: https://www.npmjs.com/package/opencode-alibaba-qwen3-auth - might help!

  9. grigio commented on Nov 2, 2025

    @grigio

    I can run gpt-oss-20b-mxfp4 locally but qwen3-coder-30b-a3b it's infinite to reply, any way to debug it?

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

Metadata

Metadata

Assignees

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions