DocumentationReference

Support Matrix

View as Markdown

This matrix defines the supported OSS targets for NeMo Platform. It covers the Python package, CLI, SDK, local services started by nemo setup, and self-managed Kubernetes deployments installed with the NeMo Platform Helm chart.

Host Platforms

AreaSupportedNotes
Deployment modenemo setup local install; Helm chart on Kubernetesnemo setup starts local services, registers a model provider, and configures the CLI and SDK. The Helm chart is the supported path for self-managed Kubernetes clusters.
Python3.12, 3.13Use an isolated virtual environment. Python 3.12 is the lowest supported Python version and Python 3.13 is the highest supported Python version.
LinuxUbuntu 22.04 LTS, Ubuntu 24.04 LTS, RHEL 9, Rocky Linux 9, Debian 12Supported for CLI, SDK, local services, and NVIDIA GPU workloads.
macOSmacOS Tahoe 26 and macOS Sequoia 15Supported for CLI, SDK, local services, and cloud/provider workflows. Local NVIDIA GPU workloads are not supported on macOS.
Architecturex86_64 Linux; Apple Silicon and Intel macOSNVIDIA GPU workloads require x86_64 Linux.
Windows and WSLNot supported for the OSS local install pathUse a supported Linux or macOS host for the local install path.
Kubernetes clustersminikube, kind, EKS, AKS, GKE, OKE, OpenShift, and on-prem KubernetesSee Install NeMo Platform with Helm for prerequisites and cluster-specific notes.

Local Runtime

AreaSupportedNotes
Package installationuv>=0.9.14,<0.10.0 recommended; pip supportedThe one-line installer sets up uv automatically. Published packages can also be installed with pip install "nemo-platform[all]".
ShellBash or zshUsed by the installer, setup commands, and documented examples.
BrowserCurrent Chrome, Edge, Firefox, or SafariRequired for the Studio UI and generated documentation.

GPU Support

AreaSupportedNotes
CUDACUDA 13 or laterGPU workloads require CUDA 13 or later.
NVIDIA driverCUDA 13-capable driver, R580 branch or laterSee the CUDA 13.0 release notes for exact driver minimums by operating system.
GPU memory40 GB VRAM minimum; 80 GB VRAM recommendedSmaller GPUs can run provider or client workflows, but local model and GPU-accelerated jobs require larger GPUs.
Recommended GPUsNVIDIA A100 80GB, H100 80GB, B200 180GBUse comparable data center GPUs that meet the CUDA and memory requirements for the model or job.
GPU platformLinux x86_64macOS and Windows are not supported for local NVIDIA GPU workloads.

Resource Profiles

ProfileCPU and memoryStorageGPU
CLI, SDK, and hosted-provider setup8 GB RAM minimum16 GB free disk spaceNot required
Local GPU workloads16 GB RAM minimum; 32 GB recommendedAdditional tens to hundreds of GB for model artifacts and job outputs40 GB VRAM minimum; 80 GB recommended

Model Providers

ProviderSupportNotes
NVIDIA BuildSupportedConfigure with NVIDIA_API_KEY.
OpenAISupportedConfigure with OPENAI_API_KEY.
AnthropicSupportedConfigure with ANTHROPIC_API_KEY.
Google GeminiSupportedConfigure with GEMINI_API_KEY.
OllamaSupported for local provider workflowsOllama must be installed and running before setup.
Custom OpenAI-compatible endpointSupportedConfigure NEMO_DEFAULT_INFERENCE_BASE_URL and NEMO_DEFAULT_INFERENCE_KEY.

Out of Scope

The following are not part of the OSS support matrix:

  • Native Windows local install.
  • Docker Compose deployment guides.
  • WSL-based local install or GPU validation.
  • macOS local GPU workloads.
  • Python earlier than 3.12 or Python 3.14 and later.
  • CUDA versions earlier than 13 for GPU workloads.