Back to Arena

Vespa AI

by Yahoo / Vespa.ai (independent OSS project)

System Card

OrganizationYahoo / Vespa.ai (independent OSS project)
Released2017-09
Architecturevector-rag / hybrid lexical-vector engine
DetailsVespa is a full-featured search and vector database engine supporting HNSW approximate nearest-neighbor search combined with BM25 lexical scoring in a single query. It runs as a stateful distributed service with multi-phase ranking, custom ranking expressions, and native ML inference. Designed for large-scale production workloads at Yahoo, it handles both batch updates and real-time writes.
Parameters—
Domainrag-retrieval
Open SourceYes
WebsiteVisit
hybrid-searchANNrankingreal-timedistributed

Capability Profile

Benchmark Scores

5 of 14 benchmarks
Long-Context Retrieval
2/5
RULER
66.614p
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
60.56p
HotpotQA
62.119p
Agent Task Memory
0/1
AgentBench-Mem
no data
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
1/1
RAGAS
67.740p
Sources:Vespa AI vendor documentation; evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)Vespa AI vendor documentation; evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)Vespa AI vendor documentation; evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)Vespa AI vendor documentation; evaluated on RAGAS: Automated Evaluation of Retrieval-Augmented Generation (Exploding Gradients, 2309)Vespa AI vendor documentation; evaluated on RULER: What's the Real Context Size of Your Long-Context Language Models (NVIDIA, 2404)