The Thinking LayerSignal & WeightPart 3quantization

Numbers from the published config.json and the FP8 release of Qwen3.8-Max. Explained in Quantization: the dial that decides how many GPUs you need.