Company intelligence
Qdrant
Composable high-performance vector search
About Qdrant
Powering the next generation of AI applications with advanced and high-performant vector similarity search technology. Qdrant is an open-source vector search engine. It deploys as an API service providing a search for the nearest high-dimensional vectors. With Qdrant, embeddings or neural network encoders can be turned into full-fledged applications for matching, searching, recommending, and much more. Make the most of your Unstructured Data!
Verified activity
Signals from Qdrant
19 published signals
Research & Knowledge
Qdrant published an article exploring how to systematically tune retrieval without guessing.
Reported by Qdrant
Products & Services
Qdrant is compatible with VectorAI DB and allows users to move to Qdrant with near-zero code changes.
Reported by Andre Zayarni
Presence & Recognition
Qdrant is hosting Vector Space Stream on September 17, a 4+ hour virtual stream focused on technical research in vector search.
Reported by Qdrant
Presence & Recognition
Qdrant is hosting Discord Office Hours on September 16, featuring open conversations with the Qdrant team.
Reported by Qdrant
Presence & Recognition
Qdrant is hosting Bring Your Own Demo Night in Berlin on September 17, featuring six community members showcasing their vector, embedding, or agent work.
Reported by Qdrant
Research & Knowledge
Qdrant published a comparison of embedded vector databases with Actian in a post.
Reported by Andre Zayarni
Products & Services
Qdrant released Qdrant 1.19.1, a patch release featuring quantized HNSW search improvements, TurboQuant SIMD implementation rework, batched HNSW searches, and payload-heavy shard transfer optimizations.
Reported by Qdrant
Presence & Recognition
Qdrant's Engineering, Research & DevRel teams are hosting a research broadcast with 8 talks on Sept 17.
Reported by Maddie D.
Presence & Recognition
Qdrant team will discuss their approaches, challenges, and learnings at Vector Space Stream in 2 weeks.
Reported by Neil Kanungo
Presence & Recognition
Qdrant published a write-up detailing the release of 3 massive datasets and an open-source tool.
Reported by Neil Kanungo
Products & Services
Qdrant released 3 massive datasets built on real-world embeddings and ground-truth queries, along with an open-source tool for producing more datasets.
Reported by Neil Kanungo
Products & Services
Qdrant released Coyo-Vector-Embeddings, a dataset for multimodal search with image-caption pairs projected through a 2048-dimensional vision-language encoder.
Reported by Andre Zayarni
Products & Services
Qdrant released PubMed-Multi-Vector, a dataset for hybrid retrieval with dense, sparse, and ColBERT-style multi-vector representations over the same corpus, 8.37 billion multi-vector tokens.
Reported by Andre Zayarni
Products & Services
Qdrant released Qdrant-FineWeb-10B, a 10 billion document dataset with dense and sparse vectors, over 50 TB of vector and meta data, and exact top-1000 ground truth computed by brute force across the full 10-billion-vector space for 120,000 queries.
Reported by Andre Zayarni
Products & Services
Qdrant released Supernova, a fully open source framework for generating ground truth, embedding generation, GPU-native brute-force ground truth, ingestion, and stress testing, running on your own infrastructure through SkyPilot.
Reported by Andre Zayarni
Research & Knowledge
Qdrant tested switching from k=2 to k=61 on a test set and saw a 42% increase in the top hybrid search result.
Reported by Andre Zayarni
Research & Knowledge
Qdrant calculates hybrid search results using k=2 and k=61 as tunable parameters, with the paper citing RRF using 60.
Reported by Andre Zayarni
Research & Knowledge
Qdrant uses k=2 as its default hybrid search tunable parameter, with the paper citing RRF using 60 as the default.
Reported by Andre Zayarni
Research & Knowledge
Qdrant tested RRF vs. DBSF distribution-based fusion method on five datasets, finding that RRF only sees rank and DBSF keeps the score gap.
Reported by Andre Zayarni