The "story" behind Kuzu v0.12.0 (now rebranded as ) is one of transition and rebranding rather than just a performance update. In October 2025, it was reported that
Prior versions relied on heap allocation via malloc. For large graphs (50GB+), this caused swapping thrashing and OOM (Out Of Memory) killer events. kuzu v0 120 better
I need to gather information about Kuzu's features, especially what's new in version 0.120. Since the user provided the original query and the example answer, I should check if Kuzu is a known company or product. Maybe it's related to graph databases or AI, given the mention of graph AI models in the example. Kuzu is a graph database system developed by Khefri, so version 0.120 probably includes new features in their graph processing or machine learning integration. The "story" behind Kuzu v0
The new Hybrid Store gives you the best of both worlds: Heap Allocation Prior versions relied on heap allocation
In previous versions, filters applied in a WHERE clause were sometimes evaluated later in the pipeline than necessary. In v0.12.0, the optimizer is now much smarter about "pushing" these filters down to the earliest possible step—often directly to the table scan.
Conclusion
Kùzu v0.1.0 (released in November 2023) represented a significant performance leap for this embedded graph database, primarily focusing on storage efficiency and query flexibility.