Qwen3.5-9B GGUF targets reasoning, tool-use

This release bridges the gap between small, fast local models and the complex reasoning capabilities typically reserved for larger frontier models.

• –Hybrid fine-tuning on Opus-4.6 Reasoning and Google's Mobile-Actions datasets significantly improves structured output reliability
• –9B parameter count provides a "sweet spot" for 8GB VRAM consumer GPUs while maintaining high instruction-following accuracy
• –Native GGUF support ensures immediate compatibility with the local LLM ecosystem without custom implementation
• –The focus on "action-oriented" prompting makes it an ideal candidate for local autonomous agents and home automation tasks
• –Quantized at Q4_K_M (5.6GB), it fits into almost any modern development environment with minimal overhead