LLM Inference

W4A8 pipeline: asymmetric 4-bit weights and dynamic rowwise FP8 activations enter a Hopper mixed-input GEMM

Is 4-Bit Quantization Truly Lossless?

Four bits are a storage budget, not a quality guarantee. Here is how asymmetric weights, MSE clipping, dynamic FP8 activations, and a purpose-built runtime make Tahoma’s W4A8 path both careful and fast.