OpenCode with Qwen3-Coder-Next with llama.cpp and LiteLLM

[nachoUC3M]

Description

OpenCode 1.2.24

Provider openai

baseURL=http://Litellm-server:4000/v1

Modelo unsloth/Qwen3-Coder-Next

  * llama.cpp -> litellm in order to response with SSE

curl /v1/models works

curl /v1/responses wors

curl /v1/responses with "stream": true return event:/data:

In OpenCode appear text part not found

Plugins

No response

OpenCode version

No response

Steps to reproduce

1. Use local Qwen3 as model
2. Write a task
3. Nothing response after many seconds thinking.

Screenshot and/or share link

No response

Operating System

No response

Terminal

No response

[github-actions]

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

• Error "reasoning part reasoning-0 not found" from tool calls using GLM 4.5 Air #6486: Similar "X part not found" error when using local models via OpenAI-compatible endpoints
• OpenCode v1.0.25 Hangs Indefinitely with LM Studio + Qwen Models Due to Empty tool_calls Array #4255: Qwen models hanging with local LLM provider (LM Studio + Qwen) via OpenAI-compatible setup
• Agent stops after tool execution with OpenAI-compatible providers (Gemini, LiteLLM) #14972: Agent stops after tool execution with OpenAI-compatible providers

If your issue is distinct, please clarify how it differs.