turbopuffer
Serverless vector and full-text search on object storage, cheap when cold
Overview
turbopuffer is a serverless search database for vectors and full text built on object storage, with BM25, filtering, multi-tenancy at huge namespace counts, native embedding and clients for Python, TypeScript, Go, Java and Ruby; regions include Singapore. Usage-based from a $16 monthly minimum; Scale $256; Enterprise from $4,096.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 1 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Storage ≤$0.33/GB, writes ≤$2/GB, queries from $1/PB
- All database features and multi-tenancy
- SOC 2 report and GDPR DPA
- Community Slack and email support
- HIPAA BAA
- SSO and audit logs
- IP allowlisting
- Private Slack channel
- Single tenancy
- Bring your own cloud
- Private networking and CMEK
- 99.95% uptime SLA, 24/7 support
Use cases
What you can produce with turbopuffer
- Serverless vector search on object storage
- BM25 full-text and hybrid search
- Multi-tenancy across huge namespace counts
- Native embedding and late interaction
- Python, TypeScript, Go, Java and Ruby clients
- BYOC and single-tenant enterprise deployment
turbopuffer has become a go-to search backend for AI companies with large, multi-tenant data, thanks to object-storage economics and a strong customer list, and its Singapore region helps ASEAN teams. Cold queries are much slower than warm ones, its cost comparisons are its own, and some widely repeated figures — including a removed Cursor case study — are out of date. Excellent for big, spiky, multi-tenant workloads; latency-critical cold paths need testing.
What we found
turbopuffer is a serverless search database for vectors and full text, built on object storage so that rarely queried data costs little to keep. It combines vector search with BM25 full-text search, filtering, multi-tenancy across very large numbers of namespaces and, since September 2026, native embedding and late-interaction pricing, with clients for Python, TypeScript, Go, Java and Ruby. Data lives in object storage and is cached on fast disks and memory when queried: its docs give a median of 14 milliseconds for warm queries on a million documents, below 10 milliseconds only for eventually consistent reads, and roughly 400 to 900 milliseconds for the first, cold query. The company calls itself ten times cheaper than alternatives, a claim of its own, and has announced a third-generation storage engine.
turbopuffer Inc. was founded in 2023 by Simon Hørup Eskildsen and Justine Li, who worked together on Shopify's infrastructure for eight years. BetaKit reported in December 2025 that the company is based in Ottawa with offices in Vancouver and New York, had raised under $1 million of primary capital before a seed round from Thrive Capital and Lachy Groom, and runs on revenue; outside estimates of its revenue are not confirmed by the company. Its customers include Anthropic, Notion, Atlassian, Linear, Ramp, Harvey, Superhuman and The New York Times; Notion credits it with an 80 per cent cost cut that let it drop per-user AI charges, according to turbopuffer's case study. Public regions include Singapore, Seoul, Sydney and Mumbai.
There is no free tier. Pricing is usage-based above a monthly minimum: Launch has a $16 minimum, cut from $64 in June 2026, with storage up to $0.33 per gigabyte, writes up to $2 per gigabyte and queries from $1 per petabyte scanned, plus SOC 2 and GDPR terms; Scale has a $256 minimum and adds HIPAA, SSO and audit logs; and Enterprise starts at $4,096 a month with single tenancy, bring-your-own-cloud and a 99.95 per cent SLA.
turbopuffer suits teams with large, multi-tenant search workloads.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including turbopuffer's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with turbopuffer unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to turbopuffer directly →
Spotted something out of date? Suggest an update →
Alternatives to turbopuffer
More in Code & Dev Tools
