Qdrant
Qdrant 1.20 adds turbo8 and turbo16 datatypes and per-element nested indexing
Qdrant 1.20 introduces two quantization-native vector datatypes, turbo8 and turbo16, along with per-element indexing for nested payload fields. According to Qdrant, the…
Qdrant Tests Cohere Embed 5 for PDF Search With One Vector per Page
Qdrant says Cohere's new Embed 5 Pro model, which represents a whole PDF page image with a single dense vector, scored above several late-interaction models on five…
Qdrant 1.20 adds BM25 on full-text indexes and TurboQuant datatypes
Qdrant 1.20.0 lets users rank full-text results with BM25 straight from payload text, with no embeddings or sparse vectors to generate. The release also brings a code…
Qdrant Tests UltraBERT-Embed: 32K-Token Vectors Don't Kill Chunking Yet
Qdrant tested whether DMIS Lab's UltraBERT-Embed-150M, which embeds up to 32K tokens in a single vector, makes document chunking obsolete. Its answer: not yet, as…
Reachy Mini robot gets a local semantic memory with Qdrant Edge
Qdrant has described how it gave Reachy Mini, an open-source robot from Pollen Robotics and Hugging Face, a semantic memory that is stored and searched on the robot…
Qdrant Shrinks Perplexity's Visual Retriever Vectors 21x With TurboQuant
Qdrant says its TurboQuant compression cuts the per-page vector size of Perplexity's pplx-embed-v2-late from 856 KB to 40 KB, a 21x reduction, while retaining 96.5% of…
Qdrant Tests Jev Classifier on Reranking, Diversity and Query Understanding
Qdrant has published a set of experiments using Jev, a decision model from TypeSafe served through OpenRouter, as a general-purpose classifier for search. According to…
Qdrant Tests How Far EmbeddingGemma 2 Vectors Can Be Shrunk
Qdrant says quantized full-size EmbeddingGemma 2 vectors kept 99% of their retrieval quality while using 30x less vector RAM. The vector database company tested Google's…
Qdrant's Rust cross-encoder library is only 1.1x faster than Python on small models
Qdrant has rewritten cross-encoder inference in Rust as a library called cross-encode-rs, and its own benchmark shows a modest gain: about 1.1x faster at the median than…