awesome-vector-databases

A curated list of vector database solutions, libraries, and resources for AI applications - https://vectordb.works
https://github.com/ever-works/awesome-vector-databases

Last synced: 6 days ago
JSON representation

Machine Learning Models
- OpenAI’s text-embedding-ada-002 - A pre-trained model used for extracting embeddings from content like PDFs, videos, and transcripts, which are then stored in vector databases for faster search. ([Read more](/details/openais-text-embedding-ada-002.md)) `embeddings` `AI` `OpenAI`
- Jina Embeddings v4 - Universal multimodal embedding model from Jina AI supporting text and images through unified pathway. Built on Qwen2.5-VL-3B-Instruct, outperforms proprietary models on visually rich document retrieval. This is a commercial API with free tier, though OSS weights available. ([Read more](/details/jina-embeddings-v4.md)) `Commercial` `Multimodal` `Open Source`
- Nomic Embed Text - First fully reproducible open-source text embedding model with 8,192 context length. v2 introduces Mixture-of-Experts architecture for multilingual embeddings. Outperforms OpenAI models on benchmarks. This is an OSS model under Apache 2.0 license. ([Read more](/details/nomic-embed-text.md)) `Open Source` `Embedding` `Multilingual`
- Cohere Embed v3 - Commercial text embedding model from Cohere with multilingual support and 1,024-dimensional vectors. Optimized for semantic search and retrieval tasks. This is a commercial API service with pay-per-use pricing. ([Read more](/details/cohere-embed-v3.md)) `Commercial` `Embedding` `Api`
- E5 Embeddings - Open-source text embedding models from Microsoft supporting 100+ languages. Features small, base, and large variants with weakly-supervised contrastive pre-training. This is an OSS model family released by Microsoft Research. ([Read more](/details/e5-embeddings.md)) `Open Source` `Microsoft` `Multilingual`
- Voyage AI Embeddings - Commercial embedding models built for enterprise-grade semantic search and RAG applications. Features voyage-3 and voyage-3-large models with multimodal support. This is a commercial API service with usage-based pricing. ([Read more](/details/voyage-ai-embeddings.md)) `Commercial` `Embedding` `Multimodal`
- BGE-M3 - A versatile embedding model from BAAI that simultaneously supports dense retrieval, sparse retrieval, and multi-vector retrieval, with multilingual support for 100+ languages and multi-granularity processing from short sentences to 8192-token documents. ([Read more](/details/bge-m3.md)) `Embedding Model` `Hybrid Search` `Multilingual`
- BGE-VL - State-of-the-art multimodal embedding model from BAAI supporting text-to-image, image-to-text, and compositional visual search. Trained on the MegaPairs dataset with over 26 million retrieval triplets. ([Read more](/details/bge-vl.md)) `Multimodal` `Open Source` `Visual Search`
- ColBERTv2 - Advanced multi-vector retrieval model creating token-level embeddings with late interaction mechanism, featuring denoised supervision and improved memory efficiency over original ColBERT. ([Read more](/details/colbertv2.md)) `Late Interaction` `Embeddings` `Retrieval`
- EmbeddingGemma - Google's 308M parameter multilingual text embedding model based on Gemma 3 that runs in less than 200MB RAM with quantization, generates embeddings in under 22ms on EdgeTPU, and ranks highest on MTEB for models under 500M parameters. ([Read more](/details/embeddinggemma.md)) `Embedding Model` `On Device` `Multilingual`
- NV-Embed - NVIDIA's generalist embedding model achieving record 69.32 score on MTEB benchmark. Fine-tuned from Llama architecture with improved techniques for training LLMs as embedding models. ([Read more](/details/nv-embed.md)) `Embeddings` `Nvidia` `Llm`
- pinecone-sparse-english-v0 - Learned sparse embedding model built on DeepImpact architecture, outperforming BM25 by up to 44% on TREC benchmarks for high-precision keyword search and hybrid retrieval. ([Read more](/details/pinecone-sparse-english-v0.md)) `Sparse` `Embeddings` `Hybrid Search`
- Qwen3 Embedding - Multilingual embedding model supporting over 100 languages and ranking #1 on MTEB multilingual leaderboard. Offers flexible model sizes from 0.6B to 8B parameters with user-defined instructions. ([Read more](/details/qwen3-embedding.md)) `Multilingual` `Open Source` `Embeddings`
- voyage-3-large - State-of-the-art general-purpose and multilingual embedding model from Voyage AI that ranks first across eight domains spanning 100 datasets, outperforming OpenAI and Cohere models by significant margins. ([Read more](/details/voyage-3-large.md)) `Embeddings` `Multilingual` `Api`
- BGE Reranker Base - Open-source cross-encoder reranking model from BAAI that enhances RAG retrieval quality by examining query-document pairs individually. Self-hostable with Apache 2.0 licensing for cost-effective production deployments. ([Read more](/details/bge-reranker-base.md)) `Reranking` `Open Source` `Rag`
- BGE-reranker-v2-m3 - Open-source multilingual reranking model from BAAI supporting 100+ languages with Apache 2.0 licensing, matching Cohere's latency on GPU with zero ongoing costs for production deployments. ([Read more](/details/bge-reranker-v2-m3.md)) `Reranking` `Multilingual` `Open Source`
- CLIP (Contrastive Language-Image Pre-training) - OpenAI's multimodal neural network trained on 400 million image-text pairs, enabling zero-shot image classification and cross-modal retrieval by learning joint embeddings for images and text. ([Read more](/details/clip-contrastive-language-image-pre-training.md)) `Multimodal` `Vision` `Openai`
- ColBERT - State-of-the-art late interaction retrieval model that produces multi-vector token-level representations, enabling efficient and effective passage search with rich contextual understanding. ([Read more](/details/colbert.md)) `Retrieval` `Multi Vector` `Neural Search`
- ColPali - Vision Language Model trained to produce high-quality multi-vector embeddings from document page images for efficient retrieval, eliminating need for OCR pipelines with ColBERT-style late interaction. ([Read more](/details/colpali.md)) `Multimodal` `Document Retrieval` `Vision`
- ColQwen - Late interaction retrieval model that applies the ColBERT token-level embedding approach using the Qwen language model as the base encoder. Provides high-quality semantic search with detailed token-level matching for improved retrieval accuracy. ([Read more](/details/colqwen.md)) `Late Interaction` `Token Level` `Semantic Search`
- ColQwen2 - A visual document retrieval model based on Qwen2-VL-2B that generates ColBERT-style multi-vector representations, treating documents as images to capture layout, tables, charts, and visual elements without requiring OCR or text extraction. ([Read more](/details/colqwen2.md)) `Visual Retrieval` `Multimodal` `Document Ai`
- E5-Mistral-7B-Instruct - Open-source embeddings model from Microsoft initialized from Mistral-7B-v0.1, achieving state-of-the-art BEIR score of 56.9 for English text embedding and retrieval tasks with 4096-dimensional vectors. ([Read more](/details/e5-mistral-7b-instruct.md)) `Embeddings` `Open Source` `Instruction Based`
- Elastic Learned Sparse Encoder - Elasticsearch's learned sparse encoding model (ELSER) that combines the efficiency of traditional search with semantic understanding. Uses neural methods to expand documents and queries with related terms while maintaining sparse representations for efficient retrieval. ([Read more](/details/elastic-learned-sparse-encoder.md)) `Sparse Encoding` `Semantic Search` `Elasticsearch`
- Gemini Embedding 2 - Google's first natively multimodal embedding model that maps text, images, video, audio and documents into a single embedding space. Supports over 100 languages with flexible output dimensions using Matryoshka Representation Learning. ([Read more](/details/gemini-embedding-2.md)) `Multimodal` `Embeddings` `Google`
- GTE Embeddings - General Text Embeddings from Alibaba DAMO Academy trained on large-scale relevance pairs. Available in three sizes (large, base, small) with GTE-v1.5 supporting 8192 context length. ([Read more](/details/gte-embeddings.md)) `Embeddings` `Open Source` `Multilingual`
- gte-Qwen2-1.5B-instruct - A state-of-the-art multilingual text embedding model from Alibaba's GTE (General Text Embedding) series, built on the Qwen2-1.5B LLM. The model supports up to 8192 tokens and incorporates bidirectional attention mechanisms for enhanced contextual understanding across diverse domains. ([Read more](/details/gte-qwen2-1-5b-instruct.md)) `Embeddings` `Multilingual` `Instruction Based` `Open Source`
- gte-Qwen2-7B-instruct - A large-scale multilingual text embedding model from Alibaba's GTE series with 7 billion parameters. Built on Qwen2-7B, it achieved a score of 70.24 on MTEB, outperforming NV-Embed-v1 and supporting 100+ languages with up to 8192 token context. ([Read more](/details/gte-qwen2-7b-instruct.md)) `Embeddings` `Multilingual` `Instruction Based` `large-model`
- ImageBind - Meta's groundbreaking multimodal embedding model that learns a joint embedding space across six modalities (images, text, audio, depth, thermal, IMU) using only image-paired data, enabling cross-modal retrieval and zero-shot capabilities. ([Read more](/details/imagebind.md)) `Multimodal` `Embedding` `Zero Shot`
- INSTRUCTOR - A task-specific text embedding model that generates customized embeddings based on natural language instructions. INSTRUCTOR achieves state-of-the-art performance on 70 diverse embedding tasks by allowing users to specify the task objective and domain. ([Read more](/details/instructor-embeddings-model.md)) `Embeddings` `Instruction Based` `Task Specific` `Open Source`
- Jina ColBERT v2 - Groundbreaking multilingual information retrieval model supporting 89 languages with token-level embeddings and late interaction. Features Matryoshka embeddings for flexible efficiency-precision tradeoffs and 8192 token input context. ([Read more](/details/jina-colbert-v2.md)) `Embedding` `Multilingual` `Colbert`
- Jina-CLIP v2 - A 0.9B multimodal embedding model with multilingual support for 89 languages, 512x512 image resolution, and Matryoshka representations that enable dimensional flexibility from 1024 down to 64 dimensions while maintaining strong performance. ([Read more](/details/jina-clip-v2.md)) `Multimodal` `Multilingual` `Embedding Model`
- jina-embeddings-v3 - Frontier multilingual text embedding model with 570M parameters and 8192 token-length, featuring task-specific LoRA adapters and outperforming OpenAI and Cohere embeddings on MTEB benchmark. ([Read more](/details/jina-embeddings-v3.md)) `Multilingual` `Embedding` `Open Source`
- jina-embeddings-v5 - Jina AI's latest embedding model achieving the highest multilingual performance among models under 1B parameters with 71.7 average MTEB score and 67.7 MMTEB score. ([Read more](/details/jina-embeddings-v5.md)) `Embeddings` `Multilingual` `Open Source`
- Llama-Embed-Nemotron-8B - Universal text embedding model from NVIDIA achieving state-of-the-art performance on MMTEB leaderboard, optimized for retrieval, reranking, semantic similarity, and classification with 4,096-dimensional embeddings. ([Read more](/details/llama-embed-nemotron-8b.md)) `Embeddings` `Multilingual` `Nvidia`
- Mixedbread AI - AI startup providing state-of-the-art embedding and reranking models through accessible APIs, offering both open-source and proprietary models optimized for various use cases. ([Read more](/details/mixedbread-ai.md)) `Embeddings` `Re Ranking` `Api`
- ModernBERT Embed - Open-source embedding model from Nomic AI based on ModernBERT-base with 149M parameters. Supports 8192 token sequences and Matryoshka Representation Learning for 3x memory reduction. ([Read more](/details/modernbert-embed.md)) `Open Source` `Embeddings` `Nlp`
- MS MARCO Cross-Encoder - Popular cross-encoder reranker models trained on MS MARCO dataset for semantic search, providing superior accuracy in re-ranking the top results from bi-encoder retrieval systems. ([Read more](/details/ms-marco-cross-encoder.md)) `Reranker` `Cross Encoder` `Search`
- multilingual-e5-large - Microsoft's state-of-the-art multilingual text embedding model supporting 100 languages with 1024-dimensional embeddings, trained on 1 billion multilingual text pairs for robust cross-lingual retrieval. ([Read more](/details/multilingual-e5-large.md)) `Multilingual` `Embedding` `Microsoft`
- mxbai-embed-large - State-of-the-art large embedding model from Mixedbread AI, ranked first among similar-sized models, supporting Matryoshka Representation Learning and binary quantization with 700M+ training pairs. ([Read more](/details/mxbai-embed-large.md)) `Embeddings` `Open Source` `Matryoshka`
- nomic-embed-text-v2-moe - Multilingual MoE text embedding model excelling at multilingual retrieval with SoTA performance compared to ~300M parameter models, supporting ~100 languages with Matryoshka Embeddings trained on 1.6B pairs. ([Read more](/details/nomic-embed-text-v2-moe.md)) `Embeddings` `Multilingual` `Local`
- Reranking Models - Cross-encoder models that rerank initial retrieval results for improved relevance. More accurate than bi-encoders but slower, typically applied to top-k candidates. ([Read more](/details/reranking-models.md)) `Reranking` `Cross Encoder` `Rag`
- SFR-Embedding - Salesforce's family of state-of-the-art embedding models including SFR-Embedding-Mistral for text and SFR-Embedding-Code for code retrieval. SFR-Embedding-Mistral achieved #1 on the MTEB benchmark with a 67.6 average score, surpassing OpenAI and Cohere models. ([Read more](/details/sfr-embedding-salesforce.md)) `Embeddings` `code` `Rag` `High Performance`
- Snowflake Arctic Embed - Suite of high-quality multilingual text embedding models optimized for retrieval performance, developed by Snowflake and available as open-source for commercial use. ([Read more](/details/snowflake-arctic-embed.md)) `Embeddings` `Multilingual` `Open Source`
- SPLADE - Sparse Lexical and Expansion Model using BERT for learned sparse retrieval, combining the interpretability of lexical search with the semantic power of neural models for enhanced keyword search. ([Read more](/details/splade.md)) `Sparse Vectors` `Retrieval` `Bert`
- stella_en - A family of English text embedding models distilled from state-of-the-art embedding models using a novel multi-stage distillation framework. Stella models support multiple dimensions (512 to 8192) through Matryoshka Representation Learning, offering flexible embedding sizes for different use cases. ([Read more](/details/stella-en-embedding.md)) `Embeddings` `Matryoshka` `distillation` `Open Source`
- text-embedding-3-large - OpenAI's flagship text embedding model with up to 3,072 dimensions, offering best-in-class performance and accuracy for English tasks with adjustable output sizes to optimize storage costs. ([Read more](/details/text-embedding-3-large.md)) `Openai` `Embeddings` `Api`
- text-embedding-3-small - OpenAI's improved embedding model with 1536 dimensions offering 5x price reduction compared to ada-002, supporting Matryoshka Representation Learning for flexible dimension sizing. ([Read more](/details/text-embedding-3-small.md)) `Openai` `Embeddings` `Cost Effective`
- UForm - Pocket-sized multimodal AI for content understanding across multilingual texts, images, and video. Up to 5x faster than OpenAI CLIP with quantization-aware embeddings and support for 20+ languages. ([Read more](/details/uform.md)) `Multimodal` `Embeddings` `Multilingual`
- Voyage Multimodal 3.5 - Next-generation multimodal embedding model built for retrieval over text, images, and videos, supporting Matryoshka embeddings with 4.56% higher accuracy than Cohere Embed v4 on visual document retrieval. ([Read more](/details/voyage-multimodal-35.md)) `Multimodal` `Embeddings` `Video`
- voyage-4 - Latest Voyage AI embedding model family featuring shared embedding space with MoE architecture, supporting flexible output dimensions and advanced quantization options for cost optimization. ([Read more](/details/voyage-4.md)) `Embeddings` `Multilingual` `Quantization`
- voyage-multimodal-3 - Voyage AI's first all-in-one multimodal embedding model supporting interleaved text and content-rich images including screenshots, PDFs, slide decks, tables, and figures. ([Read more](/details/voyage-multimodal-3.md)) `Multimodal` `Embeddings` `Visual Search`
- Cohere Rerank v3.5 - State-of-the-art foundational model for ranking with 4096 context length and multilingual support for 100+ languages. Offers exceptional performance on BEIR benchmarks and specialized domains including finance, e-commerce, and enterprise search. ([Read more](/details/cohere-rerank-v35.md)) `Reranker` `Multilingual` `Enterprise`
- Cohere Embed Multilingual v3 - High-performance multilingual embedding model from Cohere supporting 100+ languages with 1024 dimensions, optimized for semantic search, RAG, and cross-lingual retrieval tasks. ([Read more](/details/cohere-embed-multilingual-v3.md)) `Embeddings` `Multilingual` `Api`
- EmbeddingGemma - Google's text embedding model based on the Gemma architecture, available through Ollama and other platforms. Designed for generating high-quality embeddings for semantic search, retrieval, and various NLP tasks with efficient resource utilization. ([Read more](/details/embeddinggemma.md)) `Embeddings` `Google` `efficient`
- Jina Reranker v2 - Transformer-based cross-encoder model fine-tuned for text reranking with Flash Attention 2 architecture. Features multilingual support for 100+ languages, function-calling capabilities, code search, and 6x speedup over v1 with only 278M parameters. ([Read more](/details/jina-reranker-v2.md)) `Reranker` `Multilingual` `Cross Encoder`
- mGTE - Generalized long-context text representation and reranking models from Alibaba supporting 75 languages and context length up to 8192. Built on transformer++ encoder with RoPE and GLU for enhanced multilingual retrieval. ([Read more](/details/mgte.md)) `Multilingual` `Long Context` `Alibaba`
- Mistral Embed - State-of-the-art embedding model from Mistral AI that generates 1024-dimensional vectors for text, supporting semantic search, clustering, and retrieval-augmented generation applications. ([Read more](/details/mistral-embed.md)) `Embeddings` `Multilingual` `Api`
- Nomic Embed Text v1.5 - Multimodal embedding model with 137M parameters that outperforms OpenAI text-embedding-3-small on both short and long context tasks. Features Matryoshka Representation Learning for flexible embedding dimensions. ([Read more](/details/nomic-embed-text-v15.md)) `Multimodal` `Embeddings` `Open Source`
- Qwen3-VL-Embedding - Multimodal embedding model from Alibaba's Qwen family that processes text, images, and visual documents in a unified embedding space for cross-modal retrieval tasks. ([Read more](/details/qwen3-vl-embedding.md)) `Multimodal` `Embedding` `Vision` `Cross Modal`
- voyage-4-nano - The first open-weight embedding model from Voyage AI, freely available on Hugging Face under the Apache 2.0 license. This lightweight model is part of the Voyage 4 series with shared embedding space, ideal for local development and prototyping of AI applications requiring high-quality text embeddings. ([Read more](/details/voyage-4-nano.md)) `Open Source` `Embeddings` `Lightweight`
- RaDeR - RaDeR (Reasoning-aware Dense Retrieval) is a research model specifically trained on datasets that require reasoning, enabling it to learn how to retrieve relevant theorems and principles during intermediate reasoning steps. This approach allows the retriever to better generalize to diverse reasoning-intensive retrieval tasks. ([Read more](/details/rader.md)) `Dense Retrieval` `Reasoning Aware` `Research`
- Cohere Embed v4 - Multilingual, multimodal enterprise embedding model supporting over 100 programming languages and primary business languages with advanced quantization for cost optimization. ([Read more](/details/cohere-embed-v4.md)) `Embeddings` `multilingual` `Multimodal`
- Mistral Embed - State-of-the-art embedding model from Mistral AI that generates 1024-dimensional vectors for text, supporting semantic search, clustering, and retrieval-augmented generation applications. ([Read more](/details/mistral-embed.md)) `Embeddings` `multilingual` `api`
- mxbai-rerank-base-v2 - A 0.5B parameter reranking model by Mixedbread AI that provides an excellent balance of speed and accuracy, supporting 100+ languages and processing up to 8K tokens with reinforcement learning training for enhanced search relevance. ([Read more](/details/mxbai-rerank-base-v2.md)) `reranker` `multilingual` `Open Source`
- stella_en - A family of English text embedding models distilled from state-of-the-art embedding models using a novel multi-stage distillation framework. Stella models support multiple dimensions (512 to 8192) through Matryoshka Representation Learning, offering flexible embedding sizes for different use cases. ([Read more](/details/stella-en-embedding.md)) `Embeddings` `matryoshka` `distillation` `Open Source`
- vLLM - High-throughput and memory-efficient open-source LLM inference engine with PagedAttention, continuous batching, and support for embedding model serving. Widely adopted for production-scale AI inference. ([Read more](/details/vllm.md)) `inference` `Gpu Acceleration` `Open Source`
Managed and Serverless Vector DBs
- Zilliz Cloud - Zilliz Cloud is a serverless, cloud-hosted managed vector database powered by Milvus, with auto-sharding, scaling, pay-per-use pricing, automated backups, multi-region support, and RBAC/multi-tenancy. Designed for enterprise RAG and billion-scale production AI applications. Offers fully managed simplicity over self-hosted Milvus, with enterprise-grade features comparable to Qdrant Cloud. ([Read more](/details/zilliz-cloud.md)) `Milvus Based` `Autoscaling` `Enterprise` `Cloud Managed` `Serverless Scaling`
Managed & Serverless Vector DBs
- BagelDB - Collaborative vector database platform described as 'GitHub for AI data'. Features distributed storage, HNSW indexing, and supports private, collaborative, and public vector datasets. This is a commercial platform with open collaboration features. ([Read more](/details/bageldb.md)) `Commercial` `collaborative` `Distributed`
Managed Vector Databases
- AlloyDB - Google Cloud's fully managed, PostgreSQL-compatible database service that offers vector capabilities, leveraging the power of PostgreSQL and pgvector for AI applications. ([Read more](/details/alloydb.md)) `managed service` `PostgreSQL` `cloud`
- Azure Database for PostgreSQL - Microsoft Azure's managed service for PostgreSQL, which supports the pgvector extension, enabling robust vector database capabilities in the cloud for AI and machine learning workloads. ([Read more](/details/azure-database-for-postgresql.md)) `managed service` `cloud-native` `PostgreSQL`
- Amazon DocumentDB (with MongoDB compatibility) - An AWS document database service compatible with MongoDB, identified as a great choice for vector database needs. ([Read more](/details/amazon-documentdb-with-mongodb-compatibility.md)) `managed service` `document database` `MongoDB`
- Amazon RDS for PostgreSQL - A managed relational database service from AWS that can host PostgreSQL, including specific community versions, and is a suitable choice for deploying the pgvector extension for vector storage. ([Read more](/details/amazon-rds-for-postgresql.md)) `managed service` `cloud-native` `PostgreSQL`
- Aurora PostgreSQL-Compatible - An AWS database service compatible with PostgreSQL, identified as a great choice for vector database needs. ([Read more](/details/aurora-postgresql-compatible.md)) `managed service` `cloud-native` `PostgreSQL`
- Instaclustr for Managed Apache Cassandra 5.0 - A managed service offering Apache Cassandra 5.0, which can be utilized as a vector database for AI applications. ([Read more](/details/instaclustr-for-managed-apache-cassandra-50.md)) `managed service` `Cassandra` `NoSQL`
- Instaclustr for PostgreSQL - A managed service for PostgreSQL that includes support for pgvector, enabling PostgreSQL to function as a vector database for AI workloads. ([Read more](/details/instaclustr-for-postgresql.md)) `managed service` `PostgreSQL` `AI`
- DataRobot Vector Databases - The DataRobot vector databases feature provides FAISS-based internal vector databases and connections to external vector databases such as Pinecone, Elasticsearch, and Milvus. It supports creating and configuring vector databases, adding internal and external data sources, versioning internal and connected databases, and registering and deploying vector databases within the DataRobot AI platform to power retrieval-augmented generation and other AI use cases. ([Read more](/details/datarobot-vector-databases.md)) `vector databases` `RAG` `managed service`
- DataRobot Vector Database - DataRobot Vector Database is a managed vector store capability within the DataRobot AI Platform that allows users to create, register, deploy, and update vector databases for AI workloads, including RAG and semantic search. It integrates with NVIDIA NIM embeddings and supports both built-in and bring-your-own embeddings for building production-grade vector search solutions. ([Read more](/details/datarobot-vector-database.md)) `managed service` `RAG` `semantic search`
- DataRobot Vector Databases (GenAI) - A premium vector database capability within the DataRobot Generative AI platform that stores chunked unstructured text and their embeddings for retrieval-augmented generation (RAG). Users can create vector database objects, connect supported data sources from the DataRobot Data Registry, configure embeddings and chunking, and attach these vector databases to LLM blueprints in the playground to ground model responses in proprietary data. ([Read more](/details/datarobot-vector-databases-genai.md)) `RAG` `vector store` `enterprise`
- Qdrant Cloud - Qdrant Cloud is the fully managed, hosted deployment of the Qdrant vector database, providing scalable infrastructure, automation, and operations for production-grade vector search and AI workloads without managing servers. ([Read more](/details/qdrant-cloud.md)) `managed service` `cloud` `vector search`
- Qdrant Hybrid Cloud - Qdrant Hybrid Cloud is a deployment option for Qdrant that combines managed services with customer-controlled infrastructure, enabling flexible, secure, and compliant vector database deployments across cloud and private environments. ([Read more](/details/qdrant-hybrid-cloud.md)) `hybrid search` `enterprise` `managed service`
- Weaviate Cloud - Weaviate Cloud is the fully managed cloud deployment of the Weaviate vector database, providing a hosted environment for building and operating AI applications with scalable vector search, without managing infrastructure. ([Read more](/details/weaviate-cloud.md)) `managed service` `cloud-native` `vector search`
- Cloudflare Vectorize - Cloudflare Vectorize is a managed vector database/indexing service integrated with Cloudflare Workers AI. It stores and searches high-dimensional vector embeddings (such as text embeddings) using configurable dimensions and distance metrics like cosine and euclidean, automatically handling index optimization and regeneration when new data is inserted. ([Read more](/details/cloudflare-vectorize.md)) `managed service` `vector database` `cloud-native`
- Momento Vector Index - Serverless vector indexing service designed for real-time storage and retrieval of vector data. Developer-friendly with just 5 API calls to create complete indexes, featuring transparent pricing. This is a commercial managed service. ([Read more](/details/momento-vector-index.md)) `Commercial` `Serverless` `Real Time`
- Neon - Serverless Postgres with native pgvector support for vector embeddings and similarity search. Features instant provisioning, autoscaling, and scale-to-zero with separated compute and storage. This is a commercial managed service with free tier. ([Read more](/details/neon.md)) `Commercial` `Serverless` `Postgresql`
- Nuclia - AI Search and RAG-as-a-Service platform with semantic search capabilities. Features NucliaDB open-source database. Acquired by Progress in 2025, now part of Progress Agentic RAG. This is a commercial service with OSS core (NucliaDB). ([Read more](/details/nuclia.md)) `Commercial` `Open Source` `Rag`
- Upstash Vector - Serverless vector database with pay-per-use pricing and scale-to-zero capability. Fully managed service that scales to billions of vectors with simple per-request pricing. This is a commercial managed service. ([Read more](/details/upstash-vector.md)) `Commercial` `Serverless` `managed service`
- DashVector - Fully-managed, cloud-native vector search service from Alibaba Cloud based on the Proxima vector engine, offering horizontal scalability and instant vector updates for large-scale AI applications. ([Read more](/details/dashvector.md)) `Managed Service` `Cloud Native` `Scalable`
- Baidu VectorDB - Enterprise-level distributed vector database from Baidu Intelligent Cloud, built on the proprietary Mochow kernel, supporting up to 10 billion vectors with millions of QPS and millisecond latency. ([Read more](/details/baidu-vectordb.md)) `Cloud Native` `Distributed` `Chinese`
- Tencent Cloud VectorDB - Fully managed, enterprise-level distributed vector database from Tencent Cloud, supporting billion-scale vector search with millisecond latency and millions of QPS using the self-developed Olama engine. ([Read more](/details/tencent-cloud-vectordb.md)) `Cloud Native` `Distributed` `Chinese`
Multimodal Vector Databases
- YugabyteDB - PostgreSQL-compatible distributed SQL DB with HNSW vector search + keyword/full-text + relational/graph joins/aggs for hybrid queries. Scales ACID vector workloads for multimodal RAG. Unifies vectors with SQL unlike pure vector DBs like Qdrant. ([Read more](/details/yugabytedb.md)) `Distributed` `Sql` `Postgresql Compatible` `Acid` `Hnsw` `Hybrid Search` `Multimodal`
Multimodal Vector DBs
- NanoDB - NanoDB is a CUDA-optimized multimodal vector database supporting text and image vectors via CLIP embeddings for similarity search. Features multi-modal indexing in shared embedding space for text-to-image queries. Use cases include CV+text search and edge multimedia recommendations; GPU-accelerated alternative to text-only pgvector for vision-language tasks. ([Read more](/details/nanodb.md)) `Multi-Modal` `Vision-Language` `Fusion Search`
Multi Model & Hybrid Databases
- Couchbase - A database platform that includes vector support, aiming to enhance developer productivity with AI tools like Capella IQ. ([Read more](/details/couchbase.md)) `NoSQL` `vector data` `AI`
- SingleStoreDB (formerly MemSQL) - SingleStoreDB is an enterprise database that has supported vectors since 2017, in addition to exact keyword match, and recently announced support for additional vector indexes. ([Read more](/details/singlestoredb-formerly-memsql.md)) `enterprise` `SQL` `vector indexes`
- SurrealDB - A multi-model database that supports various data types and query languages, including capabilities for handling vector data. ([Read more](/details/surrealdb.md)) `multi-model` `vector data` `NoSQL`
- Azure Cosmos DB Vector Indexing - Native vector indexing capability in Azure Cosmos DB that supports flat, quantizedFlat, and diskANN index types for efficient vector similarity search using the VectorDistance function. It enables low-latency, high-throughput, and cost-efficient vector search directly in Cosmos DB collections, with options for brute-force exact search (flat), compressed brute-force search (quantizedFlat), and approximate nearest neighbor search (diskANN). ([Read more](/details/azure-cosmos-db-vector-indexing.md)) `vector search` `diskANN` `cloud-native`
- OpenSearch Vector Search - OpenSearch Vector Search is the vector similarity search and AI search capability within the OpenSearch engine, supporting vector indices, ingestion of embedding data, and search methods including raw vector search, semantic search, hybrid search, multimodal search, and neural sparse search. It enables building RAG and conversational search applications using either user-provided embeddings or embeddings generated automatically by OpenSearch. ([Read more](/details/opensearch-vector-search.md)) `vector search` `hybrid search` `semantic search`
- Rockset - Real-time analytics database with vector search capabilities, built on RocksDB with converged indexing. Acquired by OpenAI in 2024 to power retrieval infrastructure. This was a commercial service. ([Read more](/details/rockset.md)) `Commercial` `Real Time` `Analytics`
- Couchbase Vector Search - NoSQL database with vector search capabilities through Search Vector Indexes. Couchbase 8.0 introduces Hyperscale Vector Index for billion+ scale searches. This is a commercial database with free community edition. ([Read more](/details/couchbase-vector-search.md)) `Commercial` `Nosql` `Hybrid Search`
- StarRocks - Open-source high-performance analytical database with vector search capabilities. Features IVFPQ and HNSW indexing for approximate nearest neighbor search in v3.4+. This is an OSS database under Apache 2.0, a Linux Foundation project. ([Read more](/details/starrocks.md)) `Open Source` `Analytics` `Hybrid Search`
- FalkorDB GraphRAG - A unified knowledge graph and vector database solution built on Redis that seamlessly integrates graph traversal and vector similarity search for building advanced GenAI applications with both relational reasoning and semantic search capabilities. ([Read more](/details/falkordb-graphrag.md)) `Knowledge Graph` `Graph Database` `Graphrag`
- ApertureDB - Graph-vector database purpose-built for multimodal data, combining vector search with graph relationships for storing and managing images, videos, documents, embeddings, and metadata in a unified platform. ([Read more](/details/aperturedb.md)) `Multimodal` `Graph Database` `Vector Search`
- CozoDB - General-purpose, transactional, relational-graph-vector database that uses Datalog for queries. Embeddable but capable of handling large amounts of data and concurrency with HNSW indices for high-performance vector similarity searches. ([Read more](/details/cozodb.md)) `Graph Database` `Vector Search` `Datalog`
- Azure Cosmos DB Vector Indexing - Native vector indexing capability in Azure Cosmos DB that supports flat, quantizedFlat, and diskANN index types for efficient vector similarity search using the VectorDistance function. It enables low-latency, high-throughput, and cost-efficient vector search directly in Cosmos DB collections, with options for brute-force exact search (flat), compressed brute-force search (quantizedFlat), and approximate nearest neighbor search (diskANN). ([Read more](/details/azure-cosmos-db-vector-indexing.md)) `Vector Search` `Diskann` `Cloud Native`
- AtlasDB - Distributed, transactional key-value store developed by Palantir Technologies, designed for general-purpose data storage with high performance and horizontal scalability across multiple nodes. ([Read more](/details/atlasdb.md)) `Distributed` `Transactional` `Key Value Store`
- CozoDB - General-purpose, transactional, relational-graph-vector database that uses Datalog for queries. Embeddable but capable of handling large amounts of data and concurrency with HNSW indices for high-performance vector similarity searches. ([Read more](/details/cozodb.md)) `Graph Database` `Vector Search` `datalog`
- NebulaGraph - Open-source distributed graph database designed for super large-scale graphs with billions of vertices and trillions of edges. Outperforms Neo4j on larger datasets while providing graph database capabilities for AI applications. ([Read more](/details/nebulagraph.md)) `Graph Database` `Distributed` `Scalable`
Multi-Model & Hybrid Databases
- Apache Kvrocks - Distributed key-value NoSQL database with experimental vector similarity search. Redis-compatible with RocksDB storage engine, adding HNSW-based vector indexing for large-scale vector data management. ([Read more](/details/apache-kvrocks.md)) `Redis Compatible` `Distributed` `Vector Search`
- Memgraph - In-memory graph database with native vector search capabilities powered by USearch. Combines vector embeddings with knowledge graphs for GraphRAG, enabling semantic similarity search alongside graph traversal. ([Read more](/details/memgraph.md)) `Graph Database` `Vector Search` `In Memory`
Open Sources
- AnythingLLM - AnythingLLM is an open-source AI application that integrates with vector databases to facilitate storage and retrieval of embeddings, supporting various AI and LLM workflows. ([Read more](/details/anythingllm.md)) `open-source` `AI` `LLM` `vector database`
- Apache Arrow - Apache Arrow is a cross-language development platform for in-memory data that is commonly used to facilitate efficient integration between vector databases and machine learning frameworks. It provides a standardized format for data exchange that is useful for storing and querying high-dimensional vectors in AI applications. ([Read more](/details/apache-arrow.md)) `open-source` `in-memory` `data integration` `AI`
- arroy - Arroy is an open-source library for efficient similarity search and management of vector embeddings, useful in vector database systems. ([Read more](/details/arroy.md)) `open-source` `vector embeddings` `similarity search` `vector search`
- Bleve - Bleve is an open-source search library with experimental support for vector search, enabling hybrid search and retrieval in applications. ([Read more](/details/bleve.md)) `open-source` `search library` `hybrid search` `vector search`
- Crate - Crate is an open-source distributed SQL database with support for vector data types and vector search, suitable for AI-driven applications. ([Read more](/details/crate.md)) `open-source` `distributed` `SQL` `vector search`
- cuVS - cuVS is an open-source library from RAPIDS for fast, GPU-accelerated vector search, useful for building high-performance vector databases. ([Read more](/details/cuvs.md)) `open-source` `GPU acceleration` `vector search` `high-performance`
- DocArray - An open-source library for creating, storing, and searching multimodal data and vector embeddings, supporting AI and ML workflows. ([Read more](/details/docarray.md)) `open-source` `multimodal` `vector embeddings` `AI`
- Epsilla - Epsilla is an open-source vector database optimized for high-performance similarity search and scalable storage of vector embeddings. ([Read more](/details/epsilla.md)) `open-source` `vector database` `similarity search` `scalable`
- frugal - A platform focused on transforming AI/ML operations with transparency, control, and cost optimization, including support for vector database tasks. ([Read more](/details/frugal.md)) `open-source` `AI` `ML` `vector database`
- Havenask - Havenask is an open-source distributed search engine with support for vector search, designed for large-scale AI and search applications. ([Read more](/details/havenask.md)) `open-source` `distributed` `vector search` `AI`
- HelixDB - HelixDB is a powerful, open-source graph-vector database built in Rust, designed for intelligent data storage for Retrieval-Augmented Generation (RAG) and AI applications. It combines graph database features with vector search, making it directly relevant to AI and machine learning workflows that require vector data management. ([Read more](/details/helixdb.md)) `open-source` `graph database` `vector search` `RAG` `Rust`
- HVS (Hierarchical Graph Structure) - HVS is a graph-based index structure leveraging Voronoi diagrams for approximate nearest neighbor search in high-dimensional vector spaces. It is directly relevant to vector databases as it provides efficient similarity search capabilities for large-scale vector data. ([Read more](/details/hvs-hierarchical-graph-structure.md)) `open-source` `ANN` `graph database` `similarity search`
- InfluxDB - InfluxDB 3 OSS provides high-performance time series workloads with new support for vector data, making it suitable for AI/ML and vector search applications. Relevant as a vector-capable database. ([Read more](/details/influxdb.md)) `open-source` `vector data` `time series` `vector search`
- Jina - Jina is an open-source neural search framework that delivers cloud-native neural and vector search solutions powered by deep learning for AI applications. It is also known as Jina Search, designed for building search systems powered by vector databases, making it highly relevant for applications involving AI, semantic search, and vector data management. ([Read more](/details/jina.md)) `open-source` `neural search` `vector search` `cloud-native`
- KGraph - KGraph is an open-source library for fast approximate nearest neighbor search in high-dimensional vector spaces, applicable to vector database solutions. ([Read more](/details/kgraph.md)) `open-source` `ANN` `similarity search` `vector search`
- langchain4j - langchain4j is an open-source framework for developing LLM-powered Java applications, with built-in support for integrating vector databases as memory stores. ([Read more](/details/langchain4j.md)) `open-source` `LLM` `Java` `vector database`
- llm-app - llm-app is an open-source project that provides an AI application framework with integrated support for vector databases, enabling the development of LLM-powered solutions. ([Read more](/details/llm-app.md)) `open-source` `AI` `LLM` `vector database`
- MeiliSearch - MeiliSearch is an open-source, fast, and relevant search engine that supports vector search capabilities, making it suitable for AI applications requiring vector database functionality. ([Read more](/details/meilisearch.md)) `open-source` `search engine` `vector search` `AI`
- mem0 - mem0 is an open-source vector database focused on efficient storage and retrieval of high-dimensional embeddings for large-scale AI applications. ([Read more](/details/mem0.md)) `open-source` `vector database` `high-dimensional` `AI`
- MRPT - MRPT (Multi-Resolution Proximity Trees) is an open-source library for fast approximate nearest neighbor search in high-dimensional vector spaces, applicable to vector database backends. ([Read more](/details/mrpt.md)) `open-source` `ANN` `high-dimensional` `vector search`
- MuopDB - MuopDB is an open-source vector database designed for fast and scalable similarity search in AI applications. ([Read more](/details/muopdb.md)) `open-source` `vector database` `similarity search` `scalable`
- nanopq - nanopq is a lightweight product quantization library for efficient vector compression and similarity search, which is an important feature for vector databases that need to store and query large-scale vector data efficiently. ([Read more](/details/nanopq.md)) `open-source` `quantization` `vector compression` `similarity search`
- NGT - NGT (Neighborhood Graph and Tree) is an open-source vector search engine designed for fast and scalable approximate nearest neighbor search. ([Read more](/details/ngt.md)) `open-source` `vector search` `ANN` `scalable`
- OasysDB - OasysDB is an open-source vector database focused on efficient similarity search and management of high-dimensional data. ([Read more](/details/oasysdb.md)) `open-source` `vector database` `similarity search` `high-dimensional`
- puck - Puck is an open-source vector search engine designed for fast similarity search and retrieval of embedding vectors. ([Read more](/details/puck.md)) `open-source` `vector search` `similarity search` `embedding`
- RAFT - RAFT is a suite of GPU-accelerated libraries for data science, including support for vector search and similarity operations, often used in vector database scenarios. ([Read more](/details/raft.md)) `open-source` `GPU acceleration` `vector search` `data science`
- reor - reor is an open-source vector database solution focused on fast and scalable storage of high-dimensional vectors for AI and ML applications. ([Read more](/details/reor.md)) `open-source` `vector database` `scalable` `AI`
- sqlite-vec - sqlite-vec is an open-source extension for SQLite that adds vector data types and similarity search, enabling lightweight vector database capabilities. ([Read more](/details/sqlite-vec.md)) `open-source` `SQLite` `vector data` `similarity search`
- Valkey - Valkey is an open-source in-memory key-value data store that supports vector search operations, making it useful for AI and machine learning vector database workloads. It is also a specialized open-source vector database designed for efficient management and retrieval of high-dimensional vector data, offering advanced APIs and optimized storage for AI workloads. ([Read more](/details/valkey.md)) `open-source` `vector search` `in-memory` `AI`
- Awesome-Moviate - Awesome-Moviate is a movie search and recommendation engine demo that combines BM25 keyword search, semantic vector search, and hybrid search using Weaviate as the underlying vector database, serving as a practical example of hybrid retrieval for media content. ([Read more](/details/awesome-moviate.md)) `hybrid search` `examples` `open-source`
- Healthsearch Demo - Healthsearch is an open-source demo application that uses Weaviate as a vector database to retrieve supplement products based on user-written reviews and queries, illustrating real-world semantic product search over vector embeddings. ([Read more](/details/healthsearch-demo.md)) `semantic search` `examples` `open-source`
- VQLite - Lightweight and simple vector similarity search engine based on Google ScaNN. Provides a simple RESTful API for building vector similarity search services without the operational overhead of larger vector database solutions. ([Read more](/details/vqlite.md)) `Lightweight` `Scann` `Rest Api`
Open Source Vector Databases
- InfluxDB - InfluxDB 3 OSS is an open-source, self-hosted time-series database with vector data support for AI/ML workloads. Key features include high-ingest, vector search, and Apache 2.0 license. Ideal for RAG with time-series vectors, free self-hosting vs managed Pinecone for cost savings. ([Read more](/details/influxdb.md)) `Open Source` `self-hosted` `time-series` `Vector Search`
- llm-app - llm-app is an open-source, self-hosted framework for building LLM applications with vector database integration for embedding storage and retrieval. Key features include support for various vector stores and free licensing. Suitable for RAG prototypes, offering self-hosted cost advantages over managed services like Pinecone. ([Read more](/details/llm-app.md)) `Open Source` `self-hosted` `Llm`
Quantum-Safe Vector DBs
- ruqu - Rust crate for quantum circuit simulation and coherence assessment using min-cut gates. Integrates MWPM decoder and post-quantum signatures providing quantum-resistant security for AI safety in quantum-inspired vector computing environments. ([Read more](/details/ruqu.md)) `Open Source` `Rust` `Quantum` `Coherence` `Post-Quantum Crypto`
- RVF - RuVector Format (RVF) is a universal binary file format combining database, model, graph engine, kernel, and attestation into a deployable cognitive container. Provides quantum-resistant vector storage with post-quantum signatures, tamper-evident chains, and support for federated AI agent workflows. ([Read more](/details/rvf.md)) `File Format` `Cognitive Containers` `eBPF` `Wasm` `Post-Quantum Crypto` `Agentic Workflows` `Federated Learning`
Relational Databases
- CockroachDB - CockroachDB is a cloud-native, distributed SQL database that now supports vector data, combining traditional SQL queries with efficient vector search capabilities, ensuring data resilience, availability, scalability, and strong consistency. ([Read more](/details/cockroachdb.md)) `SQL` `vector data` `distributed`
- PostgreSQL - A powerful, open-source relational database that can be extended with modules like pgvector to support efficient storage and similarity search of vector embeddings, effectively functioning as a vector database. ([Read more](/details/postgresql.md)) `open-source` `relational database` `pgvector`
- ClickHouse Vector Search - Vector similarity search in ClickHouse using HNSW indexes for high-performance approximate nearest-neighbor (ANN) searches. Supports both exact brute-force and indexed search approaches with innovative QBit data type for query-time precision adjustment. ([Read more](/details/clickhouse-vector-search.md)) `Clickhouse` `Hnsw` `Analytics`
- TiDB Vector Search - Built-in vector search capability in TiDB, a MySQL-compatible database, enabling seamless storage and search for vectors using SQL with HNSW indexes. Eliminates the need for separate vector databases by combining operational and vector data. ([Read more](/details/tidb-vector-search.md)) `Mysql Compatible` `Hnsw` `Sql`
- Crunchy Data - PostgreSQL solutions provider offering managed PostgreSQL services for production workloads, including pgvector integration for vector search and embedding similarity queries. Acquired by Snowflake in June 2025 for an estimated $250M as part of a broader PostgreSQL consolidation trend. ([Read more](/details/crunchy-data.md)) `Postgresql` `Managed Service` `Enterprise`
research-papers-surveys
- SPANN - SPANN is a highly efficient billion-scale ANN search system using clustered HNSW indexes with dynamic partitioning for balanced load. Key features: disk-based, high recall, low latency on commodity hardware. Use cases: web-scale recommendation, image retrieval. Improves on DiskANN with better build time; competitive FAISS GPU in CPU perf. ([Read more](/details/spann.md)) `disk-ann` `clustered-hnsw` `Billion Scale`
Research Papers & Surveys
- A Brief Survey of Vector Databases - This survey paper provides an overview of the landscape, technologies, and applications of vector databases, making it a valuable resource for understanding the field. `vector databases` `survey` `applications` `technologies`
- A Comprehensive Survey on Vector Database - A comprehensive academic survey that explores the architecture, storage, retrieval techniques, and challenges associated with vector databases. It categorizes algorithmic approaches to approximate nearest neighbor search (ANNS) and discusses how vector databases can be integrated with large language models, offering valuable insights and foundational knowledge for understanding and building vector database systems. ([Read more](/details/a-comprehensive-survey-on-vector-database.md)) `vector databases` `survey` `ANNS` `architecture`
- ACL 2023 Tutorial: Retrieval-Based Language Models and Applications - This ACL 2023 tutorial reviews retrieval-based language models, which often rely on vector databases and vector search systems to retrieve relevant context. The tutorial covers methods and applications central to the use of vector databases in modern NLP systems. ([Read more](/details/acl-2023-tutorial-retrieval-based-language-models-and-applications.md)) `tutorials` `retrieval` `vector databases` `applications`
- ACORN - ACORN is a performant and predicate-agnostic search system for vector embeddings and structured data, enhancing the capability of vector databases to handle complex queries over high-dimensional data efficiently. ([Read more](/details/acorn.md)) `vector embeddings` `search system` `predicate-agnostic` `research`
- Adanns - Adanns is a framework for adaptive semantic search, focusing on efficient and scalable similarity search in high-dimensional vector spaces. Its relevance to 'Awesome Vector Databases' lies in its support for advanced vector search techniques suitable for AI and machine learning applications. ([Read more](/details/adanns.md)) `semantic search` `similarity search` `AI` `machine learning` `research`
- BANG - BANG is a billion-scale approximate nearest neighbor search system optimized for single GPU execution, enabling high-performance vector search in vector database environments at massive scale. ([Read more](/details/bang.md)) `ANN` `GPU acceleration` `high-performance` `vector search` `research`
- Cagra - Cagra provides highly parallel graph construction and approximate nearest neighbor search for GPUs, supporting large-scale vector database operations and efficient similarity search. ([Read more](/details/cagra.md)) `graph construction` `ANN` `GPU acceleration` `similarity search` `research`
- Efficient and Robust Approximate Nearest Neighbor Search Using Hierarchical Navigable Small World Graphs - This paper introduces the HNSW algorithm, which is widely adopted in vector databases and search engines for its efficient and robust performance on high-dimensional data. HNSW is foundational in powering modern vector search systems. ([Read more](/details/efficient-and-robust-approximate-nearest-neighbor-search-using-hierarchical-navigable-small-world-graphs.md)) `HNSW` `ANN` `vector search` `research`
- Graph-based Methods - A category of vector database solutions and algorithms leveraging graph-based approaches for efficient similarity search and vector indexing, which are core to many vector database implementations in AI applications. ([Read more](/details/graph-based-methods.md)) `graph database` `similarity search` `vector indexing` `AI`
- Li, Wen, et al. "Approximate nearest neighbor search on high dimensional data—experiments, analyses, and improvement." - An influential paper analyzing and improving approximate nearest neighbor search methods for high-dimensional data, highly relevant for developing and understanding vector databases. `ANN` `high-dimensional` `vector search` `research`
- OneSparse: A Unified System for Multi-index Vector Search - A unified system designed for efficient multi-index vector search, directly addressing large-scale vector database performance and scalability challenges. ([Read more](/details/onesparse-a-unified-system-for-multi-index-vector-search.md)) `vector search` `performance` `scalability` `research`
- Starling - Starling is an I/O-efficient, disk-resident graph index framework tailored for high-dimensional vector similarity search on large data segments, supporting the scalable storage and retrieval needs of vector databases. ([Read more](/details/starling.md)) `graph index` `similarity search` `scalable` `research`
- Towards Reliable Vector Database Management Systems: A Software Testing Roadmap for 2030 - An academic paper providing a comprehensive overview of the architecture, empirical defects, and future research roadmap for Vector Database Management Systems (VDBMS). This resource is directly relevant for understanding the current state and challenges in building and testing reliable vector databases. ([Read more](/details/towards-reliable-vector-database-management-systems-a-software-testing-roadmap-for-2030.md)) `vector databases` `testing` `roadmap` `reliability`
- VDBMS Architecture Overview - An overview of the architectural components common to Vector Database Management Systems (VDBMS), which are designed to efficiently store, index, and query high-dimensional vector embeddings. This provides foundational knowledge for anyone interested in the internal workings of vector databases. ([Read more](/details/vdbms-architecture-overview.md)) `research` `architecture` `vector databases` `high-dimensional`
- VDBMS Testing Research Roadmap Paper - A research paper that proposes the first structured roadmap for testing Vector Database Management Systems (VDBMS), analyzing bugs, vulnerabilities, and test challenges unique to vector databases. It provides insights and future directions for improving the reliability and robustness of vector databases. ([Read more](/details/vdbms-testing-research-roadmap-paper.md)) `research` `testing` `vector databases` `roadmap`
- VDBMS Testing Roadmap - A comprehensive research roadmap addressing the unique challenges of testing vector database management systems (VDBMS), including approaches for test input generation, oracle definition, and test evaluation tailored to vector databases. The work highlights the complexities of high-dimensional vector data, approximate search semantics, and integration with AI/LLM pipelines, making it a valuable resource for advancing reliability and trustworthiness in vector databases. ([Read more](/details/vdbms-testing-roadmap.md)) `vector databases` `testing` `roadmap` `AI`
- Vector Database Group @ NTU - A research group focused on advancing the theory and practice of vector databases, providing resources, publications, and tools related to vector database technology. ([Read more](/details/vector-database-group-ntu.md)) `research` `vector databases` `resources` `AI`
- Vector database management systems: Fundamental concepts, use-cases, and current challenges - A comprehensive research paper outlining the fundamental concepts, practical use-cases, and current challenges in the field of vector database management systems. `vector databases` `use-cases` `challenges` `survey`
- LANNS: a web-scale approximate nearest neighbor lookup system - A scalable system for approximate nearest neighbor search at web-scale, relevant for implementing and understanding vector database infrastructure for high-dimensional data. ([Read more](/details/lanns-a-web-scale-approximate-nearest-neighbor-lookup-system.md)) `ANN` `scalability` `vector search` `research`
- AiSAQ - AiSAQ is an all-in-storage approximate nearest neighbor search system that uses product quantization to enable DRAM-free vector similarity search, serving as a specialized vector search/indexing approach for large-scale information retrieval. ([Read more](/details/aisaq.md)) `ANN` `similarity search` `vector indexing`
- DET-LSH - DET-LSH is a locality-sensitive hashing scheme that introduces a dynamic encoding tree structure to accelerate approximate nearest neighbor (ANN) search in high-dimensional spaces. While it is a research algorithm rather than a production database, it directly targets the core operation behind vector databases—efficient ANN search over vector embeddings—and is relevant for designing or optimizing vector indexing components within vector database systems. ([Read more](/details/det-lsh.md)) `ANN` `hashing` `high-dimensional`
- Efficient Locality Sensitive Hashing - This work by Jingfan Meng is a comprehensive research thesis on efficient locality-sensitive hashing (LSH), covering algorithmic solutions, core primitives, and applications for approximate nearest neighbor search. It is relevant to vector databases because LSH-based indexing is a foundational technique for scalable similarity search over high-dimensional vectors, informing the design of vector indexes, retrieval engines, and similarity search modules in modern vector database systems. ([Read more](/details/efficient-locality-sensitive-hashing.md)) `ANN` `similarity search` `hashing`
- GTS - GTS is a GPU-based tree index for fast similarity search over high-dimensional vector data, providing an efficient ANN index structure that can be integrated into or used to build high-performance vector database systems. ([Read more](/details/gts.md)) `similarity search` `ANN` `GPU acceleration`
- Maze - Maze is a web-scale video deduplication system that relies on large-scale approximate nearest neighbor vector search over video embeddings to detect and remove duplicate or near-duplicate videos efficiently. While not a general-purpose vector database, it represents a specialized, production-scale application of vector search infrastructure for multimedia content management. ([Read more](/details/maze.md)) `ANN` `applications` `multimodal`
- SOAR - SOAR is a set of improved algorithms on top of ScaNN that accelerate vector search by introducing controlled redundancy and multi-cluster assignment, enabling faster approximate nearest neighbor retrieval with smaller indexes in large‑scale vector databases and search systems. ([Read more](/details/soar.md)) `ANN` `vector search` `optimization`
- CommVQ - A commutative vector quantization method for KV cache compression that reduces FP16 cache size by 87.5% with 2-bit quantization and enables 1-bit quantization, allowing LLaMA-3.1 8B to run with 128K context on a single RTX 4090 GPU. ([Read more](/details/commvq.md)) `Compression` `Quantization` `Llm Optimization`
- BatANN - Distributed disk-based approximate nearest neighbor system achieving near-linear throughput scaling. Delivers 6.21-6.49x throughput improvement over scatter-gather baseline with sub-6ms latency on 10 servers. ([Read more](/details/batann.md)) `Ann` `Distributed` `Research`
- Breaking the Storage-Compute Bottleneck in Billion-Scale ANNS - A 2025 research paper presenting a GPU-driven asynchronous I/O framework for billion-scale approximate nearest neighbor search. The system addresses the fundamental bottleneck of data movement between storage and compute in large-scale vector search. ([Read more](/details/breaking-storage-compute-bottleneck.md)) `Gpu Acceleration` `storage` `Algorithms` `Scalable`
- Curator - An efficient indexing approach for multi-tenant vector databases that handles low-selectivity filters effectively. Curator addresses the challenge of maintaining high performance when serving multiple tenants with filtered vector search queries. ([Read more](/details/curator-vector-search.md)) `Filtering` `Multi Tenant` `Indexing` `Optimization`
- d-HNSW - An efficient vector search system designed for disaggregated memory architectures. d-HNSW optimizes HNSW for environments where compute and memory are separated, typical in modern cloud and distributed systems. ([Read more](/details/d-hnsw-disaggregated-memory.md)) `Hnsw` `Distributed` `Cloud Native` `Optimization`
- Faster Maximum Inner Product Search in High Dimensions - A 2022 research paper presenting algorithms for faster MIPS (Maximum Inner Product Search) in high-dimensional spaces. MIPS is crucial for recommendation systems, neural networks, and various machine learning applications. ([Read more](/details/faster-mips-high-dimensions.md)) `Mips` `Algorithms` `High Dimensional` `Optimization`
- Filtered-DiskANN - Microsoft research extension to DiskANN algorithm that enables efficient label-based filtering during vector search, allowing precise results with metadata constraints without sacrificing performance. ([Read more](/details/filtered-diskann.md)) `Diskann` `Filtering` `Microsoft`
- FreshDiskANN - Fast and accurate graph-based ANN index for streaming similarity search, enabling real-time updates on billion-point indexes using a single machine with real-time freshness. ([Read more](/details/freshdiskann.md)) `Ann` `Graph Based` `Dynamic Updates`
- FusionANNS - An efficient CPU/GPU cooperative processing architecture for billion-scale approximate nearest neighbor search. FusionANNS achieves up to 13.1× higher QPS compared to SPANN and can handle billion-vector datasets with over 12,000 QPS while maintaining 15ms latency using only one entry-level GPU. ([Read more](/details/fusionanns-cpu-gpu-search.md)) `Gpu Acceleration` `Cpu` `Hybrid` `High Performance` `Scalable`
- Graph-Based Algorithms for Diverse Similarity Search - A 2026 research paper presenting graph-based algorithms for diverse similarity search, where results must be both similar to the query and diverse from each other. This addresses the common problem of redundant results in traditional similarity search. ([Read more](/details/graph-based-diverse-similarity-search.md)) `Graph Based` `Algorithms` `diversity` `Retrieval`
- In-Place Updates of Graph Index - A 2026 research paper on streaming approximate nearest neighbor search with in-place graph index updates. The approach enables real-time index modifications without expensive rebuilds, crucial for dynamic datasets. ([Read more](/details/in-place-updates-graph-index.md)) `Streaming` `Graph Based` `Algorithms` `Dynamic Updates`
- JAG - Joint Attribute Graphs for Filtered Nearest Neighbor Search, a research paper that addresses the challenge of combining vector similarity search with attribute filtering. JAG presents a novel index structure that efficiently handles filtered ANN queries common in real-world applications. ([Read more](/details/jag-joint-attribute-graphs.md)) `Filtering` `Graph Based` `Algorithms` `Hybrid Search`
- Leech Lattice Vector Quantization - Advanced vector quantization technique that explores the Leech lattice's optimal sphere packing properties at 24 dimensions. Delivers state-of-the-art LLM quantization performance, outperforming recent methods like Quip#, QTIP, and PVQ for extreme vector compression. ([Read more](/details/leech-lattice-vector-quantization.md)) `Quantization` `Compression` `Research`
- LIR: Late Interaction Workshop @ ECIR 2026 - The first workshop dedicated to late interaction and multi-vector retrieval methods at ECIR 2026, featuring keynote speaker Omar Khattab (ColBERT creator) and focusing on advances in token-level representations, multi-modal retrieval, and long-context search. ([Read more](/details/lir-late-interaction-workshop-ecir-2026.md)) `Workshop` `Late Interaction` `Academic`
- LLMs Meet Isolation Kernel - A research paper introducing lightweight, learning-free binary embeddings for fast retrieval. The approach uses isolation kernels to generate binary embeddings that dramatically reduce storage requirements (32× compression) while maintaining retrieval quality. ([Read more](/details/llm-isolation-kernel-binary-embeddings.md)) `Binary` `Compression` `Algorithms` `Lightweight`
- LoRANN - Low-Rank Matrix Factorization algorithm for Approximate Nearest Neighbor Search, offering competitive performance with faster query times than leading libraries at various recall levels. ([Read more](/details/lorann.md)) `Ann` `Algorithm` `Optimization`
- Maximum Inner Product is Query-Scaled Nearest Neighbor - A theoretical paper establishing the relationship between Maximum Inner Product Search and query-scaled nearest neighbor search. This connection enables applying NN techniques to MIPS problems with theoretical guarantees. ([Read more](/details/maximum-inner-product-query-scaled.md)) `Mips` `theory` `Algorithms` `nearest neighbor`
- MCGI - Manifold-Consistent Graph Indexing for billion-scale disk-resident vector search. Leverages Local Intrinsic Dimensionality to achieve 5.8x throughput improvement over DiskANN on high-dimensional datasets. ([Read more](/details/mcgi.md)) `Ann` `Research` `Disk Based`
- Monte Carlo Tree Search for Vector Indexing - Research on using Monte Carlo Tree Search algorithms for optimizing vector index construction and search strategies. Explores adaptive decision-making during graph building and query routing. ([Read more](/details/monte-carlo-tree-search-vectors.md)) `Algorithms` `Optimization` `Graph Based` `Research`
- OrchANN - A unified I/O orchestration framework for skewed out-of-core vector search that addresses the challenge of billion-scale ANN search when the dataset exceeds available memory. OrchANN optimizes I/O operations for graph-based indexes stored on disk. ([Read more](/details/orchann-io-framework.md)) `Disk Based` `Algorithms` `Optimization` `Scalable`
- PECANN - Parallel Efficient Clustering with graph-based Approximate Nearest Neighbor search, providing efficient clustering algorithms optimized for high-dimensional vector spaces. ([Read more](/details/pecann.md)) `Ann` `Clustering` `Parallel`
- PiPNN - An ultra-scalable graph-based nearest neighbor indexing algorithm that builds state-of-the-art indexes up to 11.6× faster than Vamana (DiskANN) and 12.9× faster than HNSW. PiPNN uses HashPrune, a novel online pruning algorithm that enables efficient billion-scale index construction on a single machine. ([Read more](/details/pipnn-ultra-scalable-indexing.md)) `Graph Based` `Indexing` `Algorithms` `High Performance`
- Pyramid Product Quantization - An advanced vector compression technique for approximate nearest neighbor search that improves upon traditional product quantization by using a hierarchical pyramid structure. Published in 2026, it achieves better compression ratios while maintaining search accuracy. ([Read more](/details/pyramid-product-quantization.md)) `product quantization` `Compression` `Algorithms` `Optimization`
- Re2G - Retrieve, Rerank, Generate system from IBM Research that combines neural retrieval and reranking with BART-based generation, achieving 9-34% gains over previous SOTA on the KILT leaderboard. ([Read more](/details/re2g.md)) `Reranking` `knowledge-intensive` `ibm`
- Scalable Distributed Vector Search - A research paper on accuracy-preserving index construction for distributed vector search systems. Published in 2025, it addresses the challenge of maintaining search quality while distributing vector indexes across multiple nodes. ([Read more](/details/scalable-distributed-vector-search.md)) `Distributed` `Scalable` `Algorithms` `Indexing`

Programming Languages

Python 59 C++ 25 Rust 20 Go 10 Jupyter Notebook 6 TypeScript 6 Java 6 C 5 JavaScript 4 C# 2

ecosyste.ms

Data

Tools

Indexes

Applications

Experiments

Awesome

awesome-vector-databases

Machine Learning Models

Managed and Serverless Vector DBs

Managed & Serverless Vector DBs

Managed Vector Databases

Multimodal Vector Databases

Multimodal Vector DBs

Multi Model & Hybrid Databases

Multi-Model & Hybrid Databases

Open Sources

Open Source Vector Databases

Quantum-Safe Vector DBs

Relational Databases

research-papers-surveys

Research Papers & Surveys