Reasoner is pluggable. There is no official Qwen3.6-9B. Qwen3.5-9B tool-call XML bugs are documented upstream.
Do
REASONER_BASE_URL + model id. RAM profile: Qwen3.5-9B. Quality: Bonsai-27B or Qwen3.6-27B.
Bake-off: same PicoClaw + 2dph MCP, measure tool_call success and VRAM. Do not claim 9B is better at tools without that.
Test
Docs name real HF ids only. Compose does not bake weights into the 2dph image.
## Why
Reasoner is pluggable. There is no official Qwen3.6-9B. Qwen3.5-9B tool-call XML bugs are documented upstream.
## Do
- `REASONER_BASE_URL` + model id. RAM profile: Qwen3.5-9B. Quality: Bonsai-27B or Qwen3.6-27B.
- Bake-off: same PicoClaw + 2dph MCP, measure tool_call success and VRAM. Do not claim 9B is better at tools without that.
## Test
- Docs name real HF ids only. Compose does not bake weights into the 2dph image.
Needs a machine with the models loaded (VRAM) and a live PicoClaw+MCP run. Docs already name Qwen3.5-9B vs Bonsai/Qwen3.6-27B. Do not claim 9B wins tools without that bake-off.
Needs a machine with the models loaded (VRAM) and a live PicoClaw+MCP run. Docs already name Qwen3.5-9B vs Bonsai/Qwen3.6-27B. Do not claim 9B wins tools without that bake-off.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Why
Reasoner is pluggable. There is no official Qwen3.6-9B. Qwen3.5-9B tool-call XML bugs are documented upstream.
Do
REASONER_BASE_URL+ model id. RAM profile: Qwen3.5-9B. Quality: Bonsai-27B or Qwen3.6-27B.Test
Needs a machine with the models loaded (VRAM) and a live PicoClaw+MCP run. Docs already name Qwen3.5-9B vs Bonsai/Qwen3.6-27B. Do not claim 9B wins tools without that bake-off.