Overview
Neo4j is the dominant property-graph database, used by more than 75% of the Fortune 100 graph-database deployments. It models data as nodes and relationships, allowing constant-time traversal regardless of graph size — a structural advantage over relational joins for connected-data problems like fraud rings, recommendation paths, supply chain analysis, identity resolution, and knowledge graphs. The query language Cypher, originally created for Neo4j, became the basis of the ISO GQL standard (released in 2024), which Neo4j now supports alongside Cypher.
The 2024–2025 cycle has been dominated by GraphRAG — the combination of vector retrieval with graph traversal for higher-quality grounding of LLM responses. Neo4j ships native vector search alongside Graph Data Science (GDS) algorithms, integrates with LangChain and LlamaIndex, and provides reference architectures for GraphRAG built on Neo4j AuraDB. The platform competes with TigerGraph (commercial graph), Amazon Neptune (managed graph on AWS), and increasingly with multi-model databases (PostgreSQL with Apache AGE, Spanner Graph) for less specialised workloads.
Key Features
- Native property-graph storage with index-free adjacency for constant-time traversal
- Cypher query language (ISO GQL compliant since 2024)
- ACID transactions with serializable isolation
- Vector search with HNSW indexes for embeddings (general availability since Neo4j 5.13)
- Graph Data Science library — 65+ algorithms (PageRank, Louvain, Node2Vec, Link Prediction)
- GraphRAG reference architectures with LangChain and LlamaIndex integrations
- Causal clustering with Raft consensus for HA and read scale-out
- Fabric for federated queries across multiple Neo4j databases
- Change Data Capture for event-driven downstream pipelines
- Browser and Bloom for visual exploration and analyst workflows
- SOC 2 Type II, HIPAA, ISO 27001 for AuraDB Enterprise
- Neo4j Aura available as first-party SaaS on AWS, Azure, GCP
Pricing
| Edition | Model | Cost |
|---|---|---|
| Neo4j Community Edition | Open source (GPLv3) | $0 (single instance, no clustering) |
| AuraDB Free | Hosted free tier | $0 (200K nodes, 400K relationships, 1 GB storage) |
| AuraDB Professional | Per GB-hour | From ~$0.09/GB-hour (~$65/month for 1 GB) |
| AuraDB Business Critical | Per GB-hour | From ~$0.27/GB-hour, multi-AZ HA |
| AuraDB Virtual Dedicated Cloud | Annual commit | From ~$30,000/year minimum |
| Neo4j Enterprise (self-hosted) | Per core/year | ~$36,000/core/year list, heavily negotiated |
Pricing verified May 2026 from Neo4j AuraDB public pricing. Graph Data Science library is included on AuraDB Professional and above. Enterprise self-hosted pricing requires direct quote; volume discounts of 40–70% are typical at scale.
Strengths
- Best-known graph database with the largest practitioner community
- Cypher and GQL provide an expressive declarative language for graph queries
- Graph Data Science library is the most comprehensive set of in-database graph algorithms
- Native vector search positions Neo4j as a GraphRAG platform without extra infrastructure
- Strong tooling — Browser, Bloom, Workspace — for analyst and developer workflows
- AuraDB is mature across all three major hyperscalers
Limitations
- Per-core enterprise pricing remains among the highest in the database market
- Causal clustering scales reads well but write throughput remains single-leader
- Operational complexity grows with very large graphs (above ~5 billion relationships)
- Hiring expertise is harder than for SQL databases; Cypher training is a real cost line
- Multi-region active-active is more limited than mainstream relational managed services