{"ok":true,"source":"tensorfeed.ai","lastUpdated":"2026-10-26","count":15,"databases":[{"id":"pinecone","name":"Pinecone","type":"managed","hostingOptions":["cloud"],"freeTier":"Starter: 2GB storage, 5 indexes, 2M write units and 1M read units per month, 1GB egress per month, 1 project, 2 users","startingPriceUSDMonth":20,"pricingNote":"Builder is $20/month flat and includes 10GB storage, 5M write units and 2M read units a month, with no published overage rate. Standard carries a $50/month usage minimum and Enterprise a $500/month minimum. Serverless usage on Standard: $0.33/GB-month storage, $4 to $4.50 per 1M write units, $16 to $18 per 1M read units, varying by cloud and region; Enterprise runs 50% higher on units at the same storage rate. Dedicated read nodes and BYOC on the upper tiers.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":true,"serverless":true,"maxVectorsPaid":"unlimited","openSource":false,"license":"Proprietary","released":"2021","url":"https://www.pinecone.io/pricing/","notes":"Most-deployed managed vector DB. Serverless removed instance sizing as a concern, and namespaces give cheap multi-tenancy. Founder Edo Liberty handed the CEO role to Ash Ashutosh in September 2025 and moved to chief scientist; Ashutosh has since said publicly that an acquisition is not the plan and is not on the table right now."},{"id":"turbopuffer","name":"Turbopuffer","type":"managed","hostingOptions":["cloud"],"freeTier":null,"startingPriceUSDMonth":16,"pricingNote":"Usage-based with monthly minimums: Launch $16, Scale $256 (adds SSO, audit logs, HIPAA BAA; audit log streaming is $128 extra), Enterprise from $4,096 with single tenancy, BYOC, a 99.95% SLA, and a 35% usage premium. Published rates: up to $0.33/GB-month storage, up to $2.00 per GB written (10KB minimum per write, batch discounts to 50% near 3.2MB), up to $1.00 per PB queried (1.28GB minimum, 80% discount from 32GB to 128GB and 96% beyond). No free tier.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":true,"serverless":true,"maxVectorsPaid":"unlimited","openSource":false,"license":"Proprietary","released":"2024","url":"https://turbopuffer.com/pricing","notes":"Stores vectors on object storage (S3/GCS) and caches hot data closer to compute, which is where the cost advantage at scale comes from. Cold queries are slower; warm queries compete with in-memory databases. Strong fit for many-tenant workloads with uneven traffic."},{"id":"qdrant-cloud","name":"Qdrant Cloud","type":"hybrid","hostingOptions":["cloud","self-host"],"freeTier":"Free forever single-node cluster: 0.5 vCPU, 1GB RAM, 4GB disk","startingPriceUSDMonth":null,"pricingNote":"Standard tier billed hourly on actual vCPU, memory, and storage usage, plus backup storage and paid inference tokens, rather than a fixed entry price. Premium requires a minimum spend. Hybrid Cloud and Private Cloud are quoted. The free tier includes cloud inference on selected models. Self-hosted Qdrant is free under Apache-2.0.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":true,"serverless":false,"maxVectorsPaid":"unlimited","openSource":true,"license":"Apache-2.0","released":"2021","url":"https://qdrant.tech/pricing/","notes":"Strong open-source story; Apache-2.0 with feature-parity between cloud and self-host. Native sparse vector support (not just bm25 hybrid). Rust core. Serverless is listed as coming soon, so capacity is still sized per cluster."},{"id":"weaviate-cloud","name":"Weaviate Cloud","type":"hybrid","hostingOptions":["cloud","self-host"],"freeTier":"Free plan: 100k objects, 1GB memory, 10GB disk, 1 collection with up to 3 tenants","startingPriceUSDMonth":45,"pricingNote":"Flex from $45/month pay-as-you-go on shared clusters (from $0.00465 per 1M vector dimensions, $0.12/GiB storage, $0.03036/GiB backup). Premium is a prepaid contract from $400/month: shared from $0.003875 per 1M dimensions, dedicated from $0.002718. Dimension rates vary by index type, compression, and region. Self-hosted Weaviate is free.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":true,"serverless":true,"maxVectorsPaid":"unlimited","openSource":true,"license":"BSD-3-Clause (optional features under the Weaviate License)","released":"2019","url":"https://weaviate.io/pricing","notes":"Schema-first, with first-class hybrid search and built-in modules for embeddings, reranking, and generative answers. License check: since late August 2026 the repo carries a wl directory under a separate Weaviate License, switched on with the WEAVIATE_LICENSE environment variable. With the variable unset, only BSD-3-Clause code runs."},{"id":"milvus-zilliz","name":"Milvus / Zilliz Cloud","type":"hybrid","hostingOptions":["cloud","self-host","embedded"],"freeTier":"Zilliz Cloud Free: 5GB storage, 2.5M vCUs per month, up to 5 collections","startingPriceUSDMonth":0,"pricingNote":"Standard from $0/month on serverless with usage billing and a 30-day trial, or from $126/GB-month on dedicated. Enterprise dedicated clusters from $197/month with a 99.95% SLA. Business Critical is quoted. Dedicated rates per million 768-dimension vectors per month: from $63 performance-optimized, $16 capacity-optimized, $5 tiered storage. On-demand compute jobs and BYOC available. Milvus self-host free under Apache-2.0.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":true,"serverless":true,"maxVectorsPaid":"tens of billions","openSource":true,"license":"Apache-2.0","released":"2019","url":"https://zilliz.com/pricing","notes":"Most battle-tested at very-large scale (10B+ vectors). GPU index option (CAGRA) for ultra-low-latency ANN. Embedded mode (Milvus Lite) for single-node test harnesses. Zilliz now markets the service as a Vector Lakebase, with on-demand compute jobs over external data."},{"id":"chromadb","name":"Chroma","type":"hybrid","hostingOptions":["cloud","self-host","embedded"],"freeTier":"Cloud Starter: $0/month plus usage with $5 free credit. Self-host: free under Apache-2.0","startingPriceUSDMonth":0,"pricingNote":"Chroma Cloud usage-based: $2.50 per logical GiB written, $0.33/GiB-month stored, $0.0075 per TiB queried, $0.09 per GiB returned. Team plan $250/month with $100 credit. Self-host or embedded is free.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":false,"serverless":true,"maxVectorsPaid":"unlimited","openSource":true,"license":"Apache-2.0","released":"2023","url":"https://www.trychroma.com","notes":"Developer-first ergonomics and the default in many LangChain and LlamaIndex tutorials. Chroma Cloud now supports sparse vectors combined with dense embeddings through reciprocal rank fusion, plus copy-on-write collection forking. The embedded library remains the easiest single-process option."},{"id":"pgvector","name":"pgvector (Postgres)","type":"oss","hostingOptions":["cloud","self-host","embedded"],"freeTier":"Same as your Postgres host (Supabase Free: 500MB database; Neon Free: 0.5GB per project)","startingPriceUSDMonth":0,"pricingNote":"Pricing follows your Postgres host. Supabase Pro from $25/month. Neon Launch and Scale are pay-as-you-go with no monthly minimum. AWS RDS varies.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":true,"serverless":true,"maxVectorsPaid":"limited by Postgres scaling","openSource":true,"license":"PostgreSQL","released":"2021","url":"https://github.com/pgvector/pgvector","notes":"Postgres extension. Best fit when your app already has Postgres and your vector workload is small to medium. Hybrid via tsvector + vector. HNSW + IVFFlat indexes."},{"id":"lancedb","name":"LanceDB","type":"hybrid","hostingOptions":["cloud","self-host","embedded"],"freeTier":"Open-source library and embedded mode free under Apache-2.0","startingPriceUSDMonth":null,"pricingNote":"LanceDB Cloud and Enterprise are priced through sales; no public rate card. The embedded library (in-process Rust with Python and JS bindings) is free.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":false,"serverless":true,"maxVectorsPaid":"unlimited","openSource":true,"license":"Apache-2.0","released":"2023","url":"https://lancedb.com","notes":"Columnar Lance file format on object storage. Embedded mode rivals SQLite for ergonomics: zero-config, in-process. Hybrid search pairs BM25 full-text with vectors and RRF reranking, and analytics queries run on the same table as vector search."},{"id":"mongodb-atlas","name":"MongoDB Atlas Vector Search","type":"managed","hostingOptions":["cloud"],"freeTier":"Free cluster (512MB storage), limited to 3 search or vector indexes","startingPriceUSDMonth":8,"pricingNote":"Flex clusters bill $0.011/hour and scale with load from $8 to a $30/month cap, with up to 10 indexes. Dedicated M10 from $0.08/hour. Production vector workloads are pointed at M10+ with dedicated Search Nodes, billed separately.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":true,"serverless":true,"maxVectorsPaid":"unlimited","openSource":false,"license":"Proprietary","released":"2023","url":"https://www.mongodb.com/products/platform/atlas-vector-search","notes":"Best fit when MongoDB is already your operational database. Hybrid via MongoDB Search ($search) plus $vectorSearch in one aggregation pipeline. MongoDB acquired Voyage AI in February 2025, so its embedding and reranking models plug into Atlas directly."},{"id":"vespa","name":"Vespa Cloud","type":"hybrid","hostingOptions":["cloud","self-host"],"freeTier":"Free trial","startingPriceUSDMonth":0,"pricingNote":"Usage-based on Vespa Cloud (compute + storage); rates are not published on the pricing page. Self-hosted Vespa free under Apache-2.0.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":true,"serverless":false,"maxVectorsPaid":"unlimited","openSource":true,"license":"Apache-2.0","released":"2017","url":"https://vespa.ai","notes":"Yahoo-built; powers very-large-scale search and recommendation systems (10B+ docs). Steepest learning curve in the catalog but the strongest hybrid retrieval and ranking story."},{"id":"elasticsearch","name":"Elasticsearch","type":"hybrid","hostingOptions":["cloud","self-host"],"freeTier":"Elastic Cloud free trial; Serverless vector profiles include 50GB free","startingPriceUSDMonth":null,"pricingNote":"Elasticsearch Serverless: search from $0.09 and ingest from $0.14 per VCU-hour, storage from $0.047/GB-month. Elastic Cloud Hosted is resource-based. Self-hosting is free under AGPLv3 or the Elastic License.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":true,"serverless":true,"maxVectorsPaid":"unlimited","openSource":true,"license":"AGPL-3.0 / SSPL / Elastic License 2.0","released":"2022 (vector)","url":"https://www.elastic.co/elasticsearch","notes":"Best fit when you already use Elastic for log or text search and want vectors co-located. Native HNSW with BM25 plus vector hybrid scoring and first-class RRF. Jina AI models are now sold as Elastic models and run natively through the Elastic Inference Service. Since September 2026 there is also a dedicated Elasticsearch Vector Database project type on Elastic Cloud Serverless that handles embedding, chunking, and index tuning for you, with Better Binary Quantization cutting vector memory by up to 32x."},{"id":"redis","name":"Redis","type":"hybrid","hostingOptions":["cloud","self-host"],"freeTier":"Redis Cloud Free: 30MB on a shared deployment, one database, best-effort SLA","startingPriceUSDMonth":5,"pricingNote":"Redis Cloud Essentials from $0.007/hour, about $5/month, on a shared deployment with 250MB to 100GB of RAM and SSD and a single database. Pro is dedicated, from $0.014/hour with a $200/month minimum and the first $200 free, and adds active-active replication and private connectivity. Self-hosted Redis Open Source is free.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":false,"serverless":false,"maxVectorsPaid":"limited by cluster RAM","openSource":true,"license":"AGPL-3.0 / RSALv2 / SSPLv1 (tri-licensed from Redis 8)","released":"2022 (vector)","url":"https://redis.io/docs/latest/develop/ai/search-and-query/vectors/","notes":"Vector similarity search shipped in RediSearch 2.4 in March 2022 and is now part of Redis Open Source. Three index types: FLAT for exact search, HNSW for approximate, and SVS-VAMANA (added in Redis 8.2) for compressed indexes, though Intel LVQ and LeanVec compression is not in the open-source build. FT.HYBRID pairs a text query with vector similarity in one call, and KNN queries filter on TEXT, TAG, NUMERIC, GEO, and GEOSHAPE fields. Best fit when Redis is already the cache or session store and the vector set fits in memory."},{"id":"opensearch","name":"OpenSearch","type":"oss","hostingOptions":["cloud","self-host"],"freeTier":null,"startingPriceUSDMonth":0,"pricingNote":"AWS OpenSearch Serverless bills OCU-hours plus storage. NextGen collections have no minimum and scale to zero after 10 idle minutes; Classic collections bill at least 2 OCUs (1 in dev-test mode). Vector collections cannot share OCUs with search collections. Self-host free under Apache-2.0.","hybridSearch":true,"metadataFiltering":true,"multiTenancy":true,"serverless":true,"maxVectorsPaid":"unlimited","openSource":true,"license":"Apache-2.0","released":"2022 (vector)","url":"https://opensearch.org","notes":"Apache-2.0 fork of Elasticsearch 7.10. AWS-native with serverless option and GPU-accelerated vector indexing on the managed service. Best fit for AWS-heavy stacks that want hybrid search without licensing entanglements."},{"id":"s3-vectors","name":"Amazon S3 Vectors","type":"managed","hostingOptions":["cloud"],"freeTier":null,"startingPriceUSDMonth":0,"pricingNote":"Pure usage in US East: $0.06/GB-month storage, $0.20/GB for PUTs, $2.50 per 1M queries plus a per-TB data-processed charge that falls as an index grows past 100K vectors.","hybridSearch":false,"metadataFiltering":true,"multiTenancy":true,"serverless":true,"maxVectorsPaid":"2 billion per index, 10,000 indexes per bucket","openSource":false,"license":"Proprietary","released":"2025","url":"https://aws.amazon.com/s3/features/vectors/","notes":"Vector storage built into S3, generally available since December 2025. Frequent queries return in about 100ms and infrequent ones in under a second, with up to 50 metadata keys per vector for filtering and 100 results per query. The cheap cold tier for large corpora; pair it with OpenSearch or a hot database when latency or hybrid search matters."},{"id":"cloudflare-vectorize","name":"Cloudflare Vectorize","type":"managed","hostingOptions":["cloud"],"freeTier":"Workers Free: 30M queried and 5M stored vector dimensions per month","startingPriceUSDMonth":5,"pricingNote":"Included with Workers Paid ($5/month account plan): 50M queried dimensions per month then $0.01 per 1M, and 10M stored dimensions then $0.05 per 100M. No charge for CPU, memory, index hours, or index count.","hybridSearch":false,"metadataFiltering":true,"multiTenancy":true,"serverless":true,"maxVectorsPaid":"20M per index (up to 1536 dimensions)","openSource":false,"license":"Proprietary","released":"2023","url":"https://developers.cloudflare.com/vectorize/platform/pricing/","notes":"Serverless vector index that lives next to Workers, Workers AI embeddings, and R2, so an agent can embed, store, and query without leaving one platform. Billing is purely vector count times dimensions, which makes cost easy to forecast. Limits to plan around: 1536 dimensions per vector, topK of 50 when returning metadata, and 10 metadata indexes per index. No native keyword search."}]}