Modified 2 August 2026

AI glossary

Quantization

In Swedish: Kvantisering

The process of compressing an AI model by reducing the precision of the numbers inside it. A model is quantized to Q4 (4-bit), Q8 (8-bit), or other levels to fit smaller GPU memory. Lower precision = smaller file and faster execution, but marginally lower quality.

Related terms

• Lossy AI
• Model
• VRAM
• Q4
• Q8