Neo4j 3.2 highlights: Scale, Governance, and Performance.
Neo4j boasts a native graph stack, and its entire approach is based on owning and optimizing this stack.
Benchmarking is typically controversial, but Neo4j claims its native graph stack gives it an unfair advantage when it comes to comparing performance with other graph solutions.
Neo4j started out as fully ACID and eventually loosened up to support what it calls causal consistency in a cluster environment. This essentially means reading one's own writes in sequence.
Neo4j has introduced its own query language, Cypher, which it claims is "by far the leading graph query language, and growing with the openCypher consortium."