Weaviate 1.39 adds 4-bit Rotational Quantization, which the company says cuts memory usage while maintaining strong recall, with faster imports and search improvements across the RQ family. 8-bit RQ remains the default in Weaviate Cloud and now imports 16% faster. An accompanying blog covers the SIMD work, recall at scale and a comparison with TurboQuant.
For retrieval systems where RAM dominates cost, 4-bit RQ offers a lower-memory option at a possible recall cost, while the default stays at 8-bit. The post states no specific recall numbers, memory reduction ratio or dataset sizes, so the trade-off must be measured on each workload.