RHEL 10 development on the NVIDIA DGX Spark workstation has officially entered its development preview phase. Red Hat announced this collaboration during the Red Hat Summit 2026 in Atlanta. Consequently, developers can now build and test artificial intelligence models locally on their desks.
The hardware integration utilizes the NVIDIA GB10 Grace Blackwell architecture to deliver up to 1 Petaflop of performance. Specifically, the system supports the FP4 data format and includes 128GB of unified memory. This setup allows teams to run local training and inference workloads without relying on external cloud systems. As a result, developers can prototype software faster.
Local Testing and Security
By running workloads locally, organizations can address challenges related to cybersecurity and operational costs. Meanwhile, developers can use embedded MLflow instances to perform trajectory tracing and evaluate autonomous agents. This local environment helps prevent sensitive data from crossing sovereign boundaries during testing. Consequently, businesses maintain full digital control over their proprietary models.
RHEL 10 Development Security Features
The operating system introduces post-quantum cryptography to protect critical production workloads against future decryption threats. Furthermore, it is the first enterprise Linux distribution to integrate Federal Information Processing Standards compliance for these cryptographic methods. These security measures run directly within the CPU and memory. Therefore, sensitive data remains protected during active processing phases.
Standardizing on this platform ensures consistency when transitioning artificial intelligence applications from local workstations to large-scale deployments. Notably, the RHEL 10 development environment integrates with Red Hat OpenShift to simplify hybrid cloud pipelines. This consistency helps developers avoid rewriting workflows when moving to production. Indeed, the unified operating foundation supports multiple chip architectures.
Platform Integration and Tools
Red Hat also updated its Podman tools to provide a local environment for building AI-enabled applications on computers. In addition, the company announced the general availability of Red Hat Desktop for local development. These tools work together to mimic production environments on a single workstation. Thus, engineering teams can reduce deployment friction significantly.
Future Availability
The broader Red Hat AI 3.4 platform is expected to become available later this month. Meanwhile, organizations can contact representatives to access the RHEL 10 development preview on NVIDIA hardware. This release represents a shift toward secure, hybrid infrastructure for enterprise operations.
Source: X (@redhat) — researched
