Quantization

AI Cost Management

A technique that reduces model size and inference cost by representing weights and activations with lower precision (e.g., 8-bit instead of 32-bit). Quantization can reduce latency and memory usage with minimal accuracy loss.