- Home
- Vector Databases
- Upstash Vector
Upstash Vector
California
What it is
Upstash Vector is a freemium, cloud-based serverless vector database for applications using vector search. It is designed for natural language processing, retrieval-augmented generation, recommendations, image recognition, clustering, and classification. DiskANN and FreshDiskANN support approximate nearest-neighbor search, while similarity options include Euclidean distance, cosine similarity, and dot product. Developers can use a REST API or SDKs for Python, TypeScript, Go, and PHP. Vectors can hold metadata, and filters can restrict queries to matching expressions. Hybrid indexes combine dense semantic search with sparse exact-token search. Integrations include Vercel AI SDK, Flowise, LangChain, Langflow, LlamaIndex, and LlamaParse. The Free plan is free, and Pay as You Go is $0.40 a month, billed per 100,000 requests. Limits include 3,072 maximum dimensions, 48 KB metadata, and 1 MB data per vector.
For and against
For
- Free plan available
- REST API and multiple language SDKs
- Metadata filtering for queries
- Hybrid dense and sparse search
- Integrations with major AI frameworks
Against
- Maximum dimensions listed as 3,072
- Default top-k limit is 1,000
- Enterprise pricing requires contacting the company
- Storage and bandwidth have separate terms
The verdict
Upstash Vector combines serverless vector storage, approximate search, metadata filtering, and hybrid indexes. Its free tier and multiple SDKs support experimentation, while higher plans provide larger data, namespace, and query allowances.
Who it suits
Upstash Vector suits developers building search, recommendation, NLP, or retrieval systems through APIs and SDKs. It offers a free plan for smaller workloads and paid options for larger limits.
Plans and prices
Free
10K daily queries/updates · 200M vectors × dimensions · 1,536 max dimensions · 100 namespaces
Read 3 Oct 2026 · upstash.com
Pay as You Go
Unlimited daily queries/updates · 2B vectors × dimensions · 3,072 max dimensions · 10K namespaces
Read 3 Oct 2026 · upstash.com
Fixed
1M daily queries/updates · 2B vectors × dimensions · 3,072 max dimensions · 10K namespaces
Read 3 Oct 2026 · upstash.com
Enterprise
Unlimited daily queries/updates · 100B vectors × dimensions · 5,000 max dimensions · unlimited namespaces
Read 3 Oct 2026 · upstash.com
On the record
- API and SDKs
- Upstash Vector provides a REST API and SDKs with first-class Python and TypeScript support.upstash.com · read 3 Oct 2026
- Company
- Upstash describes itself as a data platform for developers and says it is based in California.upstash.com · read 3 Oct 2026
- Enterprise security
- The Enterprise page lists SOC 2 and HIPAA, SAML SSO, access logs, dedicated infrastructure, and encryption and ACLs, and names GDPR and ISO 27001 among its compliance standards.upstash.com · read 3 Oct 2026
- Free trial
- Upstash says it can provide free trials for Pro and Enterprise plans for testing and proof-of-concept purposes; the page does not state a trial length.upstash.com · read 3 Oct 2026
- Hybrid search
- Hybrid indexes combine dense semantic search and sparse exact-token search, then fuse the results.upstash.com · read 3 Oct 2026
- Integrations
- The documentation provides integrations for Vercel AI SDK, Flowise, LangChain, Langflow, LlamaIndex, and LlamaParse.upstash.com · read 3 Oct 2026
- Metadata
- Vectors can have metadata attached, and query filters can restrict results to matching metadata expressions.upstash.com · read 3 Oct 2026
- Notable limits
- The pricing page lists a default topk limit of 1000 and a maximum of 48 KB metadata and 1 MB data per vector.upstash.com · read 3 Oct 2026
- Purpose
- Upstash Vector is a serverless vector database designed for high-performance vector search at scale, including NLP, RAG, recommendation, image recognition, clustering, and classification systems.upstash.com · read 3 Oct 2026
- Regions
- The pricing comparison lists regions including N. Virginia, AWS Ireland, AWS Iowa, and GCP, with availability varying by plan.upstash.com · read 3 Oct 2026
- Search engine
- The database uses DiskANN for approximate nearest-neighbor queries, which Upstash says provides high recall and low latency.upstash.com · read 3 Oct 2026
- Similarity functions
- It supports Euclidean distance, cosine similarity, and dot product.upstash.com · read 3 Oct 2026
- Support
- Enterprise features include professional support with a response-time SLA, a dedicated Slack channel, and a dedicated account manager.upstash.com · read 3 Oct 2026
