Some teaser results for a new quantization method we've been cooking up🧑🍳 GLM 5.2 is getting even faster
