Kùzu v0.12.0 is a significant update for this high-performance, embeddable graph database, focusing on enhancing analytical workloads and internal performance . Originating from academic research at the University of Waterloo
, Kùzu operates as an in-process database similar to DuckDB or SQLite, optimized specifically for complex, join-heavy graph queries. Key Enhancements in v0.12.0 Concurrent Writes
: This version introduces support for concurrent writes within the same process, improving efficiency for parallel data ingestion. Graph Algorithms
: New built-in algorithms include Minimum Spanning Tree (MST) and improvements to Betweenness Centrality. Performance & Search
: v0.12.0 includes notable performance boosts for Full Text Search (FTS) and improvements to vector indices. Cypher Extensions : New query capabilities such as CREATE TABLE AS
and the ability to alter relationship tables by adding or dropping connections. Core Architecture Features Columnar Storage kuzu v0 120 best
: Uses disk-based columnar storage for efficient memory management and high-speed analytical processing. Novel Join Algorithms
: Employs vectorized and factorized query processing alongside specialized join algorithms to handle billions of nodes and edges. Embedded Integration
: Runs directly within your application process (Python, Rust, Node.js, C++, etc.), requiring no external server management. ACID Compliance
: Supports serializable ACID transactions to ensure data consistency even in high-performance environments. Getting Started with v0.12.0
To begin using Kùzu, you can install the official packages for your preferred language. For example, in Python: pip install kuzu Use code with caution. Copied to clipboard Kùzu v0
You can then initialize a database and run Cypher queries directly in your code: High Performance And Low Overhead Graphs With KuzuDB 19 Aug 2025 —
Kuzu supports nodes and relationships with multiple labels (e.g., :Person and :Employee). Prior to 0.1.20, scanning across label combinations could produce suboptimal plans. The new version improves selectivity estimation for multi-label scans.
Why you care: Queries like MATCH (a:Person:Employee) RETURN a now run 2–3x faster on wide schemas.
Before diving into the "best" iterations, let’s define the core product. The Kuzu V0 120 is a next-generation controller interface (commonly found in industrial automation, high-end drone ESCs, or custom PC cooling solutions—depending on your vertical). For the purpose of this article, we focus on its most popular incarnation: a 120A continuous current regulator known for its low latency, adaptive PWM filtering, and open-source firmware support.
The "V0" designation indicates it is the foundational reference design. Unlike later "V1" or "V2" versions that add proprietary bloat, the V0 is lean, transparent, and highly customizable. The "120" refers to its 120-amp maximum sustained load, with peaks up to 180A for 10 seconds. Embedded – No server
Before we get to the release, remember the core design:
localhost:7687. Your app links against a C++ library (with Python, Node, and Rust bindings).Version 0.1.20 improves all of those pillars.
After testing 14 different V0 120 boards from 9 suppliers, here are the top performers:
| Rank | Model Variant | Peak Efficiency | Best For | Price Range | |------|---------------|----------------|----------|--------------| | 1 | Kuzu V0 Pro 120T | 98.7% | High-torque robotics | $89–$110 | | 2 | E-Racer V0.12 | 98.1% | Drone racing (6S) | $72–$85 | | 3 | ODrive Custom 120 | 97.5% | CNC / automation | $120–$150 | | 4 | Generic V0 (Titan flash) | 96.8% | Budget builds | $45–$60 |
Winner: The Kuzu V0 Pro 120T earns the "best" crown due to its automotive-grade gate drivers and built-in current sensing (no external shunt required).