Qwen's 8-step image model in GGUF for stable-diffusion.cpp: measured AD quants and the renders behind them.