# Quantization

> Reducing the bit-width of model weights (e.g.

from 16 to 4). 4-8× smaller memory footprint, 2-3× faster inference, ~1-2% quality loss.

---

Source: https://dfieldsolutions.com/en/glossary/quantization
DField Solutions — Torony köz 5-1.a, 2120 Dunakeszi, Hungary — +36 20 722 0002 — dezso@dfieldsolutions.com
Booking: see https://dfieldsolutions.com/en/contact