All Qwen models stop streaming mid-response when using OpenCode Go via GitHub Copilot BYOK #38053

Description

What happened

I'm using OpenCode Go as a BYOK provider through GitHub Copilot (Custom Language Models).

All Qwen models I tested stop generating after the first Agent step.

The response is truncated in the middle of a sentence without any error.

Example:

Finished with 1 step

Based on comparing the code between IV460...

The response stops there and never continues.

Expected behavior

The model should continue streaming until the response is complete.

Environment

• Client: GitHub Copilot Chat (VS Code)
• Connection: BYOK / Custom Language Models
• Provider: OpenCode Go
• Models affected: All Qwen models tested
• Agent mode

Additional notes

• Reproduced multiple times.
• Other model families (Claude, GPT, Gemini, etc.) work correctly with the same setup.
• No error message is displayed.
• The issue appears to be specific to Qwen models when accessed through OpenCode Go via GitHub Copilot BYOK.

Plugins

All Qwen models stop streaming mid-response when using OpenCode Go via GitHub Copilot BYOK

OpenCode version

No response

Steps to reproduce

1. Configure OpenCode Go as a BYOK provider in GitHub Copilot (Custom Language Models).
2. Select any Qwen model.
3. Open GitHub Copilot Chat in Agent mode.
4. Ask the model to analyze or modify a project.
5. Wait until the first tool step completes.

The response stops after "Finished with 1 step" and is truncated.

Screenshot and/or share link

Operating System

Windows 11 Pro 24H2

Terminal

No response