Back to Arena

LanceDB

by LanceDB Inc. (YC S22)

System Card

OrganizationLanceDB Inc. (YC S22)
Released2023-03
Architecturevector-rag / multimodal columnar lakehouse
DetailsLanceDB is built on the Lance columnar format, enabling zero-copy access to multimodal data (text, images, video, point clouds) stored directly on disk or object storage. It is embeddable (no server required) with Python/Rust/JavaScript bindings, supports ANN and full-text search, and can be scaled to petabytes. The serverless cloud offering adds collaboration and persistence without infrastructure management.
Parameters—
Domainrag-retrievallong-context
Open SourceYes
WebsiteVisit
embeddedserverlesscolumnarmultimodallakehouse

Capability Profile

Benchmark Scores

6 of 14 benchmarks
Long-Context Retrieval
4/5
RULER
73.368p
NIAH
no data
LooGLE
59.90p
LongBench
603p
∞Bench
70.39p
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
65.87p
MultiHop-RAG
no data
HotpotQA
61.416p
Agent Task Memory
0/1
AgentBench-Mem
no data
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:LanceDB vendor documentation; evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)LanceDB vendor documentation; evaluated on RULER: What's the Real Context Size of Your Long-Context Language Models (NVIDIA, 2404)LanceDB vendor documentation; evaluated on BABILong: Testing the Limits of LLMs with Long-Context Reasoning-in-a-Haystack (AIRI, 2406)LanceDB vendor documentation; evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)LanceDB vendor documentation; evaluated on InfiniteBench: Extending Long Context Evaluation Beyond 100K Tokens (Tsinghua / OpenBMB, 2402)LanceDB vendor documentation; evaluated on LooGLE: Can Long-Context Language Models Understand Long Contexts? (Peking University, 2311)