turbopuffer

Founded by Simon Hørup Eskildsen in 2023

Search billions of vectors and documents stored on S3 cheaply

Cost

Free Tier

Rating

★ People love it

Time to value

Quick Setup (< 1 hour)

You can use turbopuffer to run vector search and full-text search on massive datasets stored in object storage like S3. It handles approximate nearest neighbor search, BM25 full-text search, and hybrid combinations with sub-10ms latency at p50. You can filter by metadata, scale a single index to 256TB, and handle billions of documents without managing complex infrastructure. It costs about 10x less than traditional vector databases and is used in production by companies like Anthropic, Notion, and Atlassian.

What turbopuffer does

Run approximate nearest neighbor vector search with topk resultsIndex documents with BM25 for full-text keyword searchCombine vector and full-text search in a single hybrid queryFilter search results by metadata attributes alongside vector rankingUpsert millions of document embeddings into a namespaceScale a search namespace to hundreds of terabytes without manual shardingQuery cold namespaces stored on S3 with automatic caching warm-upBenchmark search latency at p50, p90, and p99 for production workloadsStores vector indexes directly on S3, making it about 10x cheaper than in-memory vector databasesSupports hybrid search combining vector similarity and BM25 full-text search in one queryScales a single namespace index up to 256TB with namespace shardingDelivers sub-10ms p50 query latency using memory and SSD caching in front of object storageHandles metadata filtering alongside vector and full-text queriesSupports unlimited namespaces with up to 250M+ seen in productionWrite throughput up to 10M+ writes per second globallyVector search recall@10 between 90-100%

Tutorials & Demos

Frequently asked

Zoī

Want a tailored answer?

See whether turbopuffer fits your stack.

Techbible weighs turbopuffer against what you already pay for, your team shape, and the work that's actually happening. Free to start.

More in Database Management

All tools →

Azimutt · Database explorer and analyzer

Next-Gen ERD: Design, Explore, Document and Analyze your database.

Chat2DB

Chat2DB is an AI-powered SQL client that transforms natural language into SQL queries. Supports MySQL, Redis, MongoDB, and enhances database management with Text2SQL and BI features.

Chroma

Open-source search and retrieval database for AI applications

ClickHouse

Fast open-source column-oriented database management system that allows generating analytical data reports in real-time using SQL queries

DataStax Astra

A Cloud-native database for modern applications.

Dragonfly

Discover Dragonfly - the most performant in-memory data store, engineered for extreme throughput at scale.

Fivetran

Automated data integration

Guepard

Clone any database in seconds.