NVIDIA's NVFP4 lossless compression trick saves ~6GB VRAM

A leaked NVIDIA technique exploits that adjacent block scales within a row differ by only a few units, compressing NVFP4 weights losslessly by 45-48% (about 6GB on typical models), freeing that VRAM for KV-cache while remaining bit-exact.

2026-10-16 ~ 2026-10-16 · 2 related posts

• NVFP4 weight lossless compression frees ~6GB VRAM for KV-cache — TheZachMueller · 2026-10-16
• New NVFP4 trick: compressing block scales saves 45-48% per expert matrix, bit-exact — TheZachMueller · 2026-10-16