Back to Arena

Notion AI

by Notion Labs

System Card

OrganizationNotion Labs
Released2023-02
Architecturevector-rag / Workspace-embedded RAG over personal and team pages
DetailsNotion AI embeds a RAG layer directly into the Notion workspace, enabling Q&A, summarization, and generation over all pages the user has access to. Meeting transcription added in 2024 captures audio from the mic/speakers (no bot) and stores transcripts as searchable pages. AI assistants query across all personal and team knowledge within permission boundaries.
Parameters—
Domainrag-retrievalepisodic-session
Open SourceNo
WebsiteVisit
workspacenote-takingteam-knowledgemeeting-transcriptionwriting-aiproductivity

Capability Profile

Benchmark Scores

6 of 14 benchmarks
Long-Context Retrieval
1/5
RULER
no data
NIAH
no data
LooGLE
no data
LongBench
603p
∞Bench
no data
Multi-Turn Recall
2/2
LoCoMo
58.710p
MemoryBank
58.61p
Cross-Session Memory
1/1
LongMemEval
55.62p
Multi-Hop QA
1/3
BABILong
no data
MultiHop-RAG
no data
HotpotQA
65.532p
Agent Task Memory
1/1
AgentBench-Mem
63.610p
Personalization
0/1
PerLTQA
no data
Factuality / Grounding
0/1
RAGAS
no data
Sources:Notion AI vendor documentation; evaluated on AgentBench Memory Track (Tsinghua KEG, 2308)Notion AI vendor documentation; evaluated on HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (Stanford / CMU, 1809)Notion AI vendor documentation; evaluated on LoCoMo: Long-Term Conversational Memory Benchmark (Snap Research, 2402)Notion AI vendor documentation; evaluated on LongBench: A Bilingual, Multitask Benchmark for Long Context Understanding (Tsinghua KEG, 2308)Notion AI vendor documentation; evaluated on LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory (Salesforce AI Research, 2410)Notion AI vendor documentation; evaluated on MemoryBank: Enhancing LLMs with Long-Term Memory (Sun Yat-sen University, 2305)