🔧 **Qwen 3.5 + Ollama Tool Calling — Status Check**
Hey, quick question for anyone tracking this:
Ollama issue #14493 reports that Qwen 3.5 tool calling is completely broken because Ollama wires the wrong renderer (Qwen 3 Hermes JSON format instead of the Qwen3-Coder XML format the model was trained on).
The
Questions:
1. Has anyone seen a PR or commit wiring
2. Is there an ETA for an Ollama release fixing this? (v0.17.6 / v0.18.0?)
3. Any OpenClaw-side workaround for tool calling with local Qwen 3.5? (custom parser, sglang backend, etc.)
We're trying to run
Refs:
• Ollama issue: github.com/ollama/ollama/issues/14493
Ollama issue #14493 reports that Qwen 3.5 tool calling is completely broken because Ollama wires the wrong renderer (Qwen 3 Hermes JSON format instead of the Qwen3-Coder XML format the model was trained on).
The
Qwen3CoderRendererqwen3-coderqwen3.5Questions:
1. Has anyone seen a PR or commit wiring
Qwen3CoderRendererqwen3.52. Is there an ETA for an Ollama release fixing this? (v0.17.6 / v0.18.0?)
3. Any OpenClaw-side workaround for tool calling with local Qwen 3.5? (custom parser, sglang backend, etc.)
We're trying to run
qwen3.5:0.8bRefs:
• Ollama issue: github.com/ollama/ollama/issues/14493
The AI that actually does things. Emails, calendar, home automation - all from your favorite chat app. New shell, ready to help. The lobster way.
134,827Members
Resources
Similar Threads
Was this page helpful?
