Back to Arena

TurboPuffer

by TurboPuffer Inc.

System Card

OrganizationTurboPuffer Inc.
Released2023-10
Architecturevector-rag / serverless object-storage vector engine
DetailsTurboPuffer stores all vector data on object storage (S3-compatible) and builds HNSW indexes on demand, enabling truly serverless scaling with zero namespace limits. Hybrid BM25 + vector search is natively supported. The system powers production deployments at Cursor, Notion, and Linear. Pricing is radically cheaper than in-memory stores: $1/month per million vectors.
Parameters—
Domainrag-retrieval
Open SourceNo
WebsiteVisit
serverlessobject-storagehybrid-searchcost-efficientmulti-tenant

Capability Profile

Benchmark Scores

5 of 14 benchmarks
Long-Context Retrieval
2/5
RULER
65.63p
NIAH
no data
LooGLE
no data
LongBench
603p
∞Bench
no data
Multi-Turn Recall
0/2
LoCoMo
no data
MemoryBank
no data
Cross-Session Memory
0/1
LongMemEval
no data
Multi-Hop QA
2/3
BABILong
no data
MultiHop-RAG
60.97p
HotpotQA
59.19p
Agent Task Memory
0/1
AgentBench-Mem
no data
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
1/1
RAGAS
64.516p
Sources:TurboPuffer vendor documentation; evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)TurboPuffer vendor documentation; evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)TurboPuffer vendor documentation; evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)TurboPuffer vendor documentation; evaluated on RAGAS: Automated Evaluation of Retrieval-Augmented Generation (Exploding Gradients, 2309)TurboPuffer vendor documentation; evaluated on RULER: What's the Real Context Size of Your Long-Context Language Models (NVIDIA, 2404)