Lepton AI is a Pythonic framework for building and running AI services, published on GitHub with about 2,800 stars. It is the official Python library and command-line tool for the NVIDIA DGX Cloud Lepton platform, letting developers deploy models and manage GPU workloads with simple Python code.
Lepton’s original design goal was to make shipping an AI service feel like writing a Python class: you define a Photon with your model code, and the framework handles packaging, scaling and serving. Following Lepton AI’s acquisition by NVIDIA, the library now operates the DGX Cloud Lepton platform: the lep CLI creates and manages endpoints, batch jobs, dev pods, storage and fine-tuning jobs, while the Python Client calls deployed endpoints like native functions. The project is Apache-2.0 licensed and under active development in 2026. The documentation includes deployment examples for chat models, embedding services and image generation, showing the same Photon pattern across very different workloads.
The Lepton library and CLI are free and open source under Apache-2.0. Running workloads uses NVIDIA DGX Cloud Lepton, a paid GPU cloud billed by usage, so costs depend on the GPU resources you consume.
Lepton AI is for ML engineers and Python developers who want to deploy and operate models on managed GPU infrastructure without hand-rolling serving and orchestration code.