Back to Arena
Vectara
by Vectara Inc.
System Card
OrganizationVectara Inc.
Released2022-10
Architecturevector-rag / grounded generation RAG-as-a-service
DetailsVectara coined the term "Grounded Generation" (now broadly called RAG) and provides a fully managed pipeline: ingest documents, chunk and embed internally, then serve with a dedicated retrieval API that returns citations. The platform uses its own Boomerang embedding model and a factual-consistency scorer to reduce hallucinations. Mockingbird, a task-specific LLM, was launched in 2024 specifically for RAG synthesis.
Parameters—
Domainrag-retrieval
Open SourceNo
WebsiteVisit
managed-raggrounded-generationanti-hallucinationcitationsenterprise
Capability Profile
Benchmark Scores
5 of 14 benchmarksMulti-Turn Recall0/2
LoCoMo
no dataMemoryBank
no dataCross-Session Memory0/1
LongMemEval
no dataMulti-Hop QA2/3
Agent Task Memory0/1
AgentBench-Mem
no dataPersonalization0/1
PerLTQA
no dataFactuality / Grounding1/1
Sources:Vectara vendor documentation; evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)Vectara vendor documentation; evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)Vectara vendor documentation; evaluated on MultiHop-RAG: Benchmarking Retrieval-Augmented Generation for Multi-Hop Queries (HKUST, 2401)Vectara vendor documentation; evaluated on RAGAS: Automated Evaluation of Retrieval-Augmented Generation (Exploding Gradients, 2309)Vectara vendor documentation; evaluated on RULER: What's the Real Context Size of Your Long-Context Language Models (NVIDIA, 2404)