Learn how Post-Training Quantization shrinks LLMs without retraining. Compare 8-bit vs 4-bit methods like SmoothQuant and AWQ, plus implementation tips for 2026.