# What is Vector Quantization?

Canonical URL: https://www.traeai.com/articles/639900de-0368-478d-aff0-9101f7dcc577
Original source: https://qdrant.tech/articles/what-is-vector-quantization/
Source name: Qdrant
Content type: article
Language: 英文
Score: 8.2
Reading time: 14 分钟
Published: 2024-09-25T12:29:33+00:00
Tags: 向量量化, Qdrant, HNSW, 嵌入压缩, 近似最近邻搜索

## Summary

向量量化通过压缩高维向量（如OpenAI嵌入）显著降低内存占用和搜索成本，Qdrant支持标量、乘积和二值化三种主要方法。

## Key Takeaways

- 1536维float32向量占6KB，百万级数据需GB级内存，量化可大幅压缩存储
- HNSW索引因随机读和图遍历导致高计算开销，量化缓解内存与延迟压力
- 标量量化将float32转为int8，实现75%内存节省，Qdrant提供配置接口

## Citation Guidance

When citing this item, prefer the canonical traeai article URL for the AI-readable summary and include the original source URL when discussing the underlying source material.