Kamiwaza v0.8.0 introduces Two-Node Community for DGX Spark, enabling larger-model inference without data center complexity
SILVERTHORNE, Colo., Dec. 18, 2025 /PRNewswire-PRWeb/ -- Kamiwaza AI, a leader in distributed AI orchestration and pioneers of the "AI Where Your Data Lives" approach, today announced the release of Kamiwaza v0.8.0, expanding its enterprise-scale AI orchestration platform with support for NVIDIA DGX Spark. Described as a "data center in a box," DGX Spark's full potential is unlocked with Kamiwaza v0.8.0 as it enables users to pool resources across devices seamlessly, making it easier to deploy high-performance AI workloads securely and efficiently, all without moving data, while leveraging NVIDIA's powerful new hardware.
"Kamiwaza v0.8.0 is about making high-performance AI practical at the point where data lives," said Luke Norris, CEO, Kamiwaza AI. "DGX Spark is an incredible platform, and with v0.8.0 we're enabling developers and AI practitioners to treat several DXG Sparks like a unified, orchestrated AI fabric, from local experimentation to production-grade deployments."
As models and datasets grow, enterprise bottlenecks are expanding from compute alone to memory capacity, data locality, and operational complexity. Kamiwaza v0.8.0 addresses these challenges directly by aligning orchestration, scheduling, and deployment around modern accelerated systems, starting with DGX Spark, resulting in faster iteration, stronger security and governance, and fewer "environment" headaches, all while keeping data in place.
The "Two-Node" Community Breakthrough
Today's largest frontier models frequently exceed the memory limits of a single desktop-class system, typically pushing teams toward rack-mounted servers and complex infrastructure. With Kamiwaza Community Edition v0.8.0, Kamiwaza introduces a specialized "Two-Node" mode designed for paired NVIDIA DGX Spark systems:
- Automatic pair detection: Kamiwaza detects linked DGX Spark pairs connected via high-speed interconnects.
- One logical supercomputer: The platform treats the linked pair as a single, unified resource pool.
- Intelligent model splitting: The Kamiwaza scheduler automatically manages model parallelism, splitting workloads across the pair without requiring users to manually shard layers.
- Unified memory awareness: Memory computation accounts for the paired-node configuration, enabling use of the combined 256GB+ unified memory pool to handle larger datasets and models than a single node could support.
This "Two-Node" approach allows teams to run bigger models with fewer operational tradeoffs, bringing advanced workloads into reach for developers, labs, and small teams without sacrificing performance or security.
Optimized for DGX Spark & NVIDIA Blackwell-class systems
Kamiwaza v0.8.0 adds DGX Spark-aware orchestration and deployment, with Unified Memory Architecture (UMA), making both single- and two-node configurations easier to run, tune, and operationalize.
- DGX Spark-ready resource management: Kamiwaza's resource manager and model engine are now DGX Spark-aware and UMA-aware, enabling straightforward model deployment via Kamiwaza's model engine in both single- and two-node configurations.
- DGX Spark-aware container management:: Kamiwaza v0.8.0 detects DGX Spark systems and automatically leverages the DGX Spark-optimized service container stack, including support for the sm121 instruction set, for an optimal model serving experience.
Enterprise Scale: The "One API" Promise
While Community Edition introduces a powerful two-node capability, Kamiwaza Enterprise operationalizes the same approach across clusters of n-nodes, managing entire fleets of DGX Sparks as a distributed edge cloud.
This is where Kamiwaza's "One API" promise becomes concrete:
- Seamless scalability: Kamiwaza Enterprise users can prototype locally on DGX Spark, then scale to larger clusters with orchestrated fleet management.
- True portability: Workloads developed on DGX Spark can run unchanged across production environments, delivering "write once, run anywhere" portability for AI agents and applications.
Kamiwaza v0.8.0 is available now. Complete instructions for downloading and installing Community and Enterprise Edition are available at https://docs.kamiwaza.ai/. To learn more about deploying distributed AI with Kamiwaza, visit www.kamiwaza.ai.
About Kamiwaza AI
Kamiwaza AI delivers the distributed inference platform that makes AI deployment safe, governable, and scalable across any environment—cloud, on-prem, or edge. Its orchestration engine enables intelligent agents to run where data lives, with built-in contextual authorization that ensures security and compliance from day one. By unifying model deployment, agent tooling, and enterprise governance, Kamiwaza helps organizations accelerate AI transformation while maintaining full data sovereignty and control. Learn more at https://www.kamiwaza.ai/.
Media Contact
Emily Townsend, Scratch Marketing + Media for Kamiwaza AI, 1 9788883729, [email protected], https://www.kamiwaza.ai/
SOURCE Scratch Marketing + Media for Kamiwaza AI
Share this article