Releases·Tool Release
Qdrant releases FineWeb-10B, PubMed multi-vector, and Coyo embedding datasets plus open-source Supernova benchmark tooling
Qdrant released new large-scale vector datasets, including Qdrant-FineWeb-10B on Hugging Face, to improve vector database benchmarking. A Qdrant employee shared the release on Reddit, citing frustration that existing vector DB benchmarks rely on unrealistic random vectors, small dimensions, and poor methodology. The company footed the compute cost with Vultr's support.
6 engagement1 sourceWed, Sep 2, 2026, 06:42 PM
AI Overview
- Qdrant, with Vultr compute support, released Qdrant-FineWeb-10B: 10.07B dense and 10.07B sparse vectors from FineWeb using gte-multilingual-base
- Qdrant-FineWeb-10B includes exact brute-force top-1000 ground truth for 120,000 dense, sparse, and filtered queries
- Qdrant released PubMed-Multi-Vector with BGE-M3 dense, sparse, and ColBERT-style multi-vector representations over an identical corpus for hybrid retrieval comparisons
- Qdrant released Coyo-Vector-Embeddings: 15.4M dense vectors at 2048 dims from Qwen3-VL-Embedding-2B over image-caption pairs, all free on Hugging Face
- Qdrant open-sourced Supernova, a YAML-driven framework for embedding, GPU brute-force ground-truthing, and stress testing targeting Qdrant, Milvus, and Elasticsearch
Related
Social·Community
Stop Renting Your AI's Memory — Dylan Couzon, Qdrant
Dylan Couzon, developer relations engineer at Qdrant, gave a talk at AI Engineer World's Fair 2026 arguing practitioners should own their AI's memory rather than keep it in vendor clouds. He frames model as CPU, context as RAM, and memory as disk, and argues retrieval beats stuffing context. A live fully offline demo shows a drone building searchable memory with sub-millisecond queries in a 15 MB footprint using Qdrant Edge.
Dylan Couzon, developer relations engineer at Qdrant, gave a talk at AI Engineer World's Fair 2026 arguing practitioners should own their AI's memory rather than keep it in vendor clouds. He frames model as CPU, context as RAM, and memory as disk, and argues retrieval beats stuffing context. A live fully offline demo shows a drone building searchable memory with sub-millisecond queries in a 15 MB footprint using Qdrant Edge.
71 engagement1 sourceyoutubeFri, Oct 2, 2026, 07:00 AM
Products·launched
raggio: open source hybrid vector DB with 4-bit quantization
raggio is an open-source hybrid vector database that uses TurboQuant 4-bit quantization to shrink vector indexes by about 8x, enabling large document collections on small hardware. It offloads collections to disk, so storage isn't limited by RAM, and is designed as a plug-and-play alternative to commercial hosted services or hand-rolled FAISS. It targets companies and individuals needing a self-hosted, resource-efficient vector database for production workloads.
raggio is an open-source hybrid vector database that uses TurboQuant 4-bit quantization to shrink vector indexes by about 8x, enabling large document collections on small hardware. It offloads collections to disk, so storage isn't limited by RAM, and is designed as a plug-and-play alternative to commercial hosted services or hand-rolled FAISS. It targets companies and individuals needing a self-hosted, resource-efficient vector database for production workloads.
1 engagement1 sourceredditMon, Aug 24, 2026, 03:10 PM
Products·building
TensoRAG – offline multi-domain vector compression using multilinear GSVD in NumPy
TensoRAG is an open-source hobby project implementing multi-tenant/multi-domain vector compression via multilinear GSVD, based on a 2024 TU Ilmenau PhD thesis. It runs entirely offline in NumPy with no API calls, targeting RAG builders who need compressed embeddings across multiple domains. An interactive demo and agent simulator are referenced.
TensoRAG is an open-source hobby project implementing multi-tenant/multi-domain vector compression via multilinear GSVD, based on a 2024 TU Ilmenau PhD thesis. It runs entirely offline in NumPy with no API calls, targeting RAG builders who need compressed embeddings across multiple domains. An interactive demo and agent simulator are referenced.
1 engagement1 sourceredditMon, Sep 7, 2026, 08:14 PM
Products·launched
Lance-bundle: Portable embeddings for offline vector search
Lance-bundle is a library that packages precomputed embedding vectors with the embedding model into a single file, enabling instant semantic search without re-embedding. It uses LanceDB and ONNX for low dependency footprint and fast cold start, solving the problem of repeatedly generating embeddings for local RAG applications.
Lance-bundle is a library that packages precomputed embedding vectors with the embedding model into a single file, enabling instant semantic search without re-embedding. It uses LanceDB and ONNX for low dependency footprint and fast cold start, solving the problem of repeatedly generating embeddings for local RAG applications.
3 engagement2 sourceshackernews, redditTue, Aug 11, 2026, 02:53 PM
Social·Community
Researcher seeks RAG test datasets for retrieval comparison
A university researcher posted on Reddit asking for datasets to evaluate RAG retrieval methods, specifically comparing dense, sparse, and hybrid approaches. They need collections with documents, queries, and ground-truth relevance labels.
A university researcher posted on Reddit asking for datasets to evaluate RAG retrieval methods, specifically comparing dense, sparse, and hybrid approaches. They need collections with documents, queries, and ground-truth relevance labels.
3 engagement1 sourceredditTue, Jul 28, 2026, 07:58 PM