Qwen3 Coder on OpenCode Zen: Tool call metadata being printed to chat instead of actually being called #10855

Activity

1. jeremyakers

   added

   bugSomething isn't workingSomething isn't working

   on Jan 27, 2026
2. github-actions

   assigned

   fwang

   on Jan 27, 2026
3. github-actions

   added

   zenRelates to ZenRelates to Zen

   on Jan 27, 2026
4. github-actions commented on Jan 27, 2026

   

   github-actionsBot

   on Jan 27, 2026 – with GitHub Actions

   Contributor

   More actions

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

   • qwen3-coder fails to call edit tool #6918: qwen3-coder fails to call edit tool - Same model (Qwen3 Coder) failing with invalid tool call arguments
   • Only return tool_call parameters but can not run the tool with local LLM #7486: Only return tool_call parameters but can not run the tool with local LLM - Similar symptom where tool call metadata is shown instead of being executed
   • ver 1.1.36, what leads to the interruption of tool calls, the conversation stoped quite frequently #10673: Tool call interruption with Qwen3 model - Related conversation interruption issues

   Feel free to ignore if none of these address your specific case.
5. bi-kai commented on Jan 29, 2026

   

   bi-kai

   on Jan 29, 2026

   More actions

   same issue, with GLM 4.7, this issue comes up quite frequently.

   👍React with 👍3Reacted by Mattwmaster58, Jihyeon Gim and George
6. lee-b commented on Jan 30, 2026

   

   lee-b

   on Jan 30, 2026

   More actions

   Yes, it happens with nemotron 3 nano too. I note that, if a tool call attempt is made with invalid parameters, then opencode tells the model that it made a mistake in the tool call, asks it to retry. I believe it should do the same if XML-style tool call attempts are detected.
7. Mattwmaster58 commented on Feb 1, 2026

   

   Mattwmaster58

   on Feb 1, 2026

   Last edited by Mattwmaster58

   More actions

   Getting this with GLM-4.7 as well. I was on 1.1.28 previous, now on 1.1.48. What I suspect the issue was with continuing a conversation created with an old version. I haven't been able to repro the issue since creating a new conversation
8. RawthiL commented on Feb 3, 2026

   

   RawthiL

   on Feb 3, 2026

   More actions

   any chance this is related to this vLLM issue? vllm-project/vllm#22975

   I facing this when running a local instance of vLLM and Qwen3-30B and qwen3_xml tool parset. The model just prints the xml of the tool call.
9. github-actions commented on Apr 5, 2026

   

   github-actionsBot

   on Apr 5, 2026 – with GitHub Actions

   Contributor

   More actions

   To stay organized issues are automatically closed after 90 days of no activity. If the issue is still relevant please open a new one.
10. github-actions

    closed this as completedon Apr 5, 2026