Qdrant 1.20 - Full-Text Search Improvements

Qdrant 1.20 introduces direct BM25 scoring on full-text indexes, allowing payload text to be searched and ranked without generating sparse-vector embeddings, alongside a code tokenizer that improves matching across naming conventions such as camelCase and snake_case. The release also accelerates nested payload filters through per-element indexes that can avoid payload reads, with reported benchmark gains of 7 to 38 times. New TurboQuant formats add turbo16, which halves float32 storage while retaining near-float32 accuracy, and turbo8, which further reduces storage with recall near float16 and improved search speeds; 8-bit TurboQuant quantization is also available for regular float vectors. Memory-tier controls now extend to ID trackers, while revised inline vector storage eliminates duplicate vector copies but may alter RAM requirements depending on HNSW index settings. Security changes require authentication for internal cluster communication when API keys are configured and permit internal ports to be bound to private interfaces. Additional features include GCS and Azure snapshot support, shared snapshot bucket prefixes, configurable audit-log output, collection creation timestamps, safer replica removal, local-time log timestamps, substring matching, and upgrade guidance for both Qdrant Cloud and self-hosted deployments.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.