Hi, I’m Eric. I’ve been working with AI since the Llama 1 days and I’m here to share what I know. I post my research, projects, and essays here.
Models, measurements, and thoughts
Latest writing
Qwen3.6 27B Quantization Across Engines: Size vs. KL Divergence
Sixteen Qwen3.6 27B quantizations compared by loaded weight size, KL divergence, and top-1 agreement—including GGUF under llama.cpp and several quantization formats under vLLM.