# Quantization

> Quantization shrinks a model by storing its weights at lower numerical precision (for example 8-bit instead of 16-bit), cutting memory use and speeding up inference with minimal quality loss.

*Source: https://www.lazlosoftwaresolution.com/glossary/quantization*

Quantisation is a practical lever for running capable models on cheaper hardware or at higher throughput. It is widely used when self-hosting open models to hit latency and cost targets without retraining.