NVIDIA DGX Cloud Lepton
Connect developers to a global network of GPU compute for building and deploying AI
Paid·Technical·Powered by NVIDIA·API available
Key strengths
Global GPU compute network unificationDesigned for AI-native teams and model buildersPowered by NVIDIA DGX-class hardware (Blackwell, Hopper)Supports full ML lifecycle: build, train, deployNo infrastructure management required
Paid only
Santa Clara, USA
No ratings yet
- Distributed LLM training – Launch multi-node, multi-GPU training jobs across NVLink-connected clusters using Hopper or Blackwell GPUs.
- Model serving with auto-scaling – Deploy HuggingFace or custom models as REST API endpoints with replica scaling managed by Lepton.
- MLOps pipeline integration – Connect Lepton's compute layer into existing MLOps workflows via API, enabling automated training and deployment pipelines.
- Fine-tuning with managed environments – Use pre-configured CUDA environments to fine-tune foundation models (e.g., Llama, Mistral) without environment setup overhead.
- GPU resource orchestration – Leverage NVIDIA Run:ai integration for intelligent GPU scheduling, workload prioritization, and utilization optimization across teams.