ggml-org/llama.cpp

turbo3 at 3.5 bpv competes with upstream rotated q4_0 at 4.5 bpv

turbo3 is SMALLER (3.5 bpv vs 4.5 bpv) AND BETTER than upstream rotated q4_0 at all context lengths. The gap widens with context — at 65K turbo3 degrades only +0.53% from f16 baseline while q4_0_rot degrades +1.73%. ggerganov's PR #21038 adds Hadamard rotation to standard quant types (citing TurboQuant by name), but their approach applies rotation to existing block quantization formats. Our tighter Lloyd-Max codebook + norm correction combo outperforms at lower bit rate.