{
  "metadata": {
    "name": "Vector Databases for RAG: Deployment-Shape, License, Hybrid-Search and Pricing Comparison 2026",
    "description": "A structured comparison of 11 vector databases for retrieval-augmented generation (RAG) across five dimensions: deployment shape, open-source license, native hybrid search, metadata filtering, and pricing model. Compiled vendor-neutral with conflict disclosure; each cell read from the vendor's own repository, documentation, or pricing page. The thesis: hybrid search and filtering are now table-stakes, so the deciding axis is deployment shape and data gravity, not benchmark.",
    "toolsCompared": 11,
    "deploymentShapes": [
      "Embedded",
      "Self-Hosted",
      "Managed",
      "Serverless"
    ],
    "dimensions": [
      "deployment_shape",
      "open_source_license",
      "hybrid_search",
      "metadata_filtering",
      "pricing_model"
    ],
    "datePublished": "2026-06-17",
    "dateModified": "2026-06-17",
    "license": "CC-BY-4.0",
    "licenseUrl": "https://creativecommons.org/licenses/by/4.0/",
    "canonicalUrl": "https://nesyona.com/research/vector-db-rag-comparison-2026/",
    "methodology": "Each cell read from the vendor's own public repository, documentation, or pricing page as of June 2026. Open-source licenses verified from the project's GitHub repository (Redis confirmed as the Redis 8 tri-license RSALv2/SSPLv1/AGPLv3, not a single permissive license). Hybrid search marked 'yes' only where a single native query path is documented; 'partial' denotes availability through an integration or full-text add-on. Metadata filtering is near-universal across the category. Vendor performance and cost claims are attributed, not asserted; no first-party recall or latency benchmark is encoded here. The four-shape model and the rankings were fixed before any monetization check; no paid placement.",
    "creator": "Vincent Wesley Couey",
    "orcid": "0009-0005-6869-308X",
    "publisher": "Nesyona",
    "definedTerm": "Vector Database Deployment Shapes -- a four-shape reference model (Embedded, Self-Hosted, Managed, Serverless) for choosing a RAG vector store by operational fit and data gravity rather than by recall benchmark, given that hybrid search and metadata filtering are now standard across the category.",
    "dataProvenance": "SOURCED: every matrix cell read from the vendor's own public repository, documentation, or pricing page as of June 2026; open-source licenses verified from each project's GitHub. DERIVED: none material — a sourced capability/pricing matrix."
  },
  "data": [
    {
      "tool": "pgvector",
      "shape": "Extension (in Postgres)",
      "open_source_license": "PostgreSQL License",
      "hybrid_search": "partial (Postgres FTS)",
      "metadata_filtering": "yes (SQL)",
      "pricing_model": "free (pay your Postgres)",
      "best_for": "apps already on Postgres, under ~10M vectors"
    },
    {
      "tool": "Qdrant",
      "shape": "Self-Hosted + Managed",
      "open_source_license": "Apache-2.0",
      "hybrid_search": "yes (sparse+dense)",
      "metadata_filtering": "yes (strong, Rust)",
      "pricing_model": "usage / free tier",
      "best_for": "fast filtered search at low self-host cost"
    },
    {
      "tool": "Weaviate",
      "shape": "Self-Hosted + Managed",
      "open_source_license": "BSD-3-Clause",
      "hybrid_search": "yes (native BM25)",
      "metadata_filtering": "yes",
      "pricing_model": "resource-based",
      "best_for": "native hybrid search and module ecosystem"
    },
    {
      "tool": "Milvus / Zilliz",
      "shape": "Self-Hosted + Managed",
      "open_source_license": "Apache-2.0",
      "hybrid_search": "yes (2.4+)",
      "metadata_filtering": "yes",
      "pricing_model": "free OSS / usage (Zilliz)",
      "best_for": "billion-scale distributed workloads"
    },
    {
      "tool": "Pinecone",
      "shape": "Serverless (Managed)",
      "open_source_license": "proprietary",
      "hybrid_search": "yes (sparse-dense)",
      "metadata_filtering": "yes",
      "pricing_model": "usage (serverless)",
      "best_for": "zero-ops fully-managed RAG"
    },
    {
      "tool": "Chroma",
      "shape": "Embedded + Cloud",
      "open_source_license": "Apache-2.0",
      "hybrid_search": "partial (integration)",
      "metadata_filtering": "yes",
      "pricing_model": "free OSS / cloud usage",
      "best_for": "prototyping and local-first RAG"
    },
    {
      "tool": "LanceDB",
      "shape": "Embedded + Cloud",
      "open_source_license": "Apache-2.0",
      "hybrid_search": "partial (FTS)",
      "metadata_filtering": "yes",
      "pricing_model": "free OSS / cloud",
      "best_for": "multimodal and edge / embedded"
    },
    {
      "tool": "turbopuffer",
      "shape": "Serverless (Managed)",
      "open_source_license": "proprietary",
      "hybrid_search": "yes (BM25+vector)",
      "metadata_filtering": "yes",
      "pricing_model": "usage (object-storage)",
      "best_for": "storage-heavy, cost-optimized workloads"
    },
    {
      "tool": "Vespa",
      "shape": "Self-Hosted + Managed",
      "open_source_license": "Apache-2.0",
      "hybrid_search": "yes (native ranking)",
      "metadata_filtering": "yes",
      "pricing_model": "free OSS / resource (Cloud)",
      "best_for": "complex ML-driven ranking over text+vectors"
    },
    {
      "tool": "Marqo",
      "shape": "Self-Hosted + Cloud",
      "open_source_license": "Apache-2.0",
      "hybrid_search": "yes",
      "metadata_filtering": "yes",
      "pricing_model": "free OSS / cloud",
      "best_for": "end-to-end embedding generation + storage"
    },
    {
      "tool": "Redis (Query Engine)",
      "shape": "Self-Hosted + Managed",
      "open_source_license": "RSALv2 / SSPLv1 / AGPLv3 (Redis 8 tri-license)",
      "hybrid_search": "yes (vector + text)",
      "metadata_filtering": "yes",
      "pricing_model": "free OSS / Redis Cloud",
      "best_for": "low-latency vectors where Redis is already in the stack"
    }
  ]
}