YFarmX

4-bit quantisation

AI

4-bit quantisation: Compressing model weights to just four bits each, roughly a quarter of the usual size, so large models fit on smaller or cheaper hardware.

Related terms

Browse the full glossary →