Tag
bytkim releases a 4-bit QLoRA SFT Multi-Token Prediction fine-tune of Qwen3.6-27B, packaged as GGUF for local agentic coding. The no-thinking tune is designed for low-latency direct output in agent loops.