Templates Models

Build

AI Studio Notebooks GPU VMs ComfyUI

Train

Interactive training Experiment logging Training clusters PyTorch Lightning

Inference

Model APIs Dedicated models Deploy custom models Managed inference

GPUs Solutions Docs Pricing

Model APIsAccess top AI models instantly.
40 million free tokens to start.Dedicated modelsRun models on dedicated instances with instant autoscaling.Deploy a custom modelTurn your container into a scalable API in one click. Run on any cloud.

Top AI modelsBring AI into your product today. Production-grade models that scale with you.

Nemotron 3 Ultra

MiniMax M2.5 Deepseek V4 Pro

GPT-5.5 Gemma 4

Claude Fable 5 View all

AI StudioBuild and run AI in a collaborative GPU workspace. Use any IDE via SSH.NotebooksExperiment and analyze data in persistent GPU notebooks with AI help.GPU VMsLaunch raw GPU machines with full SSH access and persistent storage.

Build with appsRun AI apps in the cloud with zero setup.

ComfyUI OpenClaw

Interactive trainingCollaborative GPU workspace where AI agents help train and optimize models.Training clustersManaged frontier-grade GPU clusters with SLURM, K8s or multi-cloud LEC.PyTorch LightningMake large-scale PyTorch training simple. Train any model of any size.

Customize your training stackFor teams that need deep control, expert optimizations, and high-performance training.

Multi-node Faster data loading Batch Jobs

Thunder LitLogger

Build and test with GPUsInteractive workspace with agents to build & debug before going live.Deploy your own modelTurn your container into a scalable API in one click. Run on any cloud.Expert-managed inferenceOptimized and operated by Lightning’s PyTorch experts.

Customize your inference stackFor teams that need deeper control, custom pipelines, or full infrastructure control.

Agents (LitAI)Lightning deploy Batch Jobs Pipelines

On-demand GPUsRun VMs with direct ssh across 8+ clouds.Managed GPU clustersReserved clusters with SLURM, K8s. Flexible terms.GPU marketplaceGet the best rates and GPU availability on 8+ clouds.

Live GPU prices

L4$0.48T4$0.19L40S$2.89A100$1.55H200$4.33B200$12.54L4$0.48T4$0.19L40S$2.89A100$1.55H200$4.33B200$12.54

EnterpriseGo from prototypes to production AI securely.Frontier labsFully managed dedicated GPU clusters. Expert support.Model optimizationsFaster training and inference by the PyTorch experts.

AcademiaAccelerate research in universities and labs.

Customer stories

All docsDocs hub for the full Lightning ecosystem.Lightning AI StudioCloud workspace where AI helps you build and ship.PyTorch LightningPretrain and finetune any model of any size.

Learn

Tutorials Blog Community Courses Events OSS program About LitLogger

Pricing overviewExplore plans, GPU rates, model API prices, and FAQs.EnterpriseCustom contracts, dedicated support, and volume pricing.Frontier labsCompetitive prices for GPU clusters and 24/7 support.

Startups

All startups YC startups NVIDIA inception NVIDIA connect

Academia

Students Classrooms Research labs Universities

Request demoLog in Start free Open app

Lightning AI

Platform

Lightning AIDevelop, train and deploy on Lightning AI Python SDKProgrammatic access to the Lightning AI platform CLIAccess the Lightning AI platform from the command line

Open Source Libraries

PyTorch LightningFinetune and pretrain AI models. Focus on science, not engineering.Lightning FabricScale foundation models to 1000s of GPUs with expert-level control TorchMetrics90+ PyTorch metrics, optimized for distributed training ThunderCompiler for PyTorch: Make PyTorch models faster!LitLoggerExperiment tracking built into Lightning AI LitDataStream and process datasets at scale LitServeEasily serve AI models — Lightning fast LitAILLM router and minimal agent framework in one LitGPT20+ high-performance LLMs with recipes to pretrain, finetune and deploy at scale.

Overview

Getting started Why Lightning AI

ML use case Startups Frontier labs Research project Enterprise Admin tools Model development

Idea phase Early stage Growth

Students Classrooms University class Research project Research projects Research labs University lab Academic institutions University Academic conferences

Students The side project Side projects The work project Work projects

Coding interviews Hackathons Workshops

Account creation GitHub vs Lightning AI Billing Get help Security Personal organization

Build

AI Studio (workspace)

Start and stop Studio Collaboration GitHub and GitLab Connect local IDE SSH access Add GPUS Deploy a Studio Add data Base Studios Modify environment Environment persistence Background execution Debug interactively Deploy on public ports Publishing Duplicate Studios Auto sleep Custom Docker images Interruptible machines Environment variables Import models Manage disk size Move clouds On-start actions On-stop actions Managed secrets GPU marketplace Plugins CLI

Run arbitrary code in a sandboxed environment Build a stateful, sandboxed code interpreter Build a coding agent with LangChain

Expose web apps Comfy UI Streamlit apps Gradio apps Host demos

Inference

Overview Model APIs

Lightning Inference

Autoscaling Clouds and regions Ideal workflow Secure communication Multi-node Inference Websockets SDK

Custom inference engine Huggingface models vLLM and others Model registry Compile models

Self-serve vLLM inference server

Submit jobs Job outputs Prototype server APIs Job pipeline (DAG)Common issues SDK CLI Examples

Deploy a container

AWS ECR DockerHub GitHub containers GCP artifact registry Lit container registry

Deploy Examples (full code)

Agent with tool use BERT model Fast API server Hugging Face models Image generation Image models - CLIP Image super resolution Image QA - Llama 3 Image QA - Phi 3 LLM - Llama 3 Media conversion Multi-modal LLM Noise cancellation Proxy server RAG - LlamaIndex Random forest Speech to text - Whisper Stable diffusion 2 TensorFlow model Text embedding Voice clone - Coqui XTTS Voice clone - SVTT2 vLLM XGBoost model

Manage execution order Pipeline environment Available node types Scheduling Versioning Shared filesystem Examples

Turn-key models

Train

Experiment Management

Installation API Reference

Standalone Usage Lightning Integration Logging Artifacts Logging Media Workflows Examples

Quickstart Files, Media, Models Lightning Tutorial Complete Workflow LitServe Tutorial

SDK commands CLI commands

Pretrain models

Multi-GPU training PyTorch Lightning Lightning Fabric Continual pretraining Hyperparameter sweeps TorchMetrics Accelerate library

Finetune models

Finetune LLMs Single GPU training LoRa finetuning Instruction finetuning RAG Proxy tuning Training guide Hyperparameter sweeps PyTorch Lightning Lightning Fabric TorchMetrics

Examples (low code)

Instruction finetune LLMs Pretrain LLM

Examples (full code)

Build a custom Trainer Finetune any LLM Hyperparameter sweep HuggingFace models Image segmentation Image classification LSTM recurrent net Music generation Object detection Pretrain any LLM Pretrain diffusion model Recurrent neural net Recommender system Text classification Text summarization Timeseries forecasting

GPU Cloud

Overview GPU marketplace

Managed Kubernetes Managed SLURM Networking Storage

Bring your own cloud

Data

Overview Upload data Move data Drive

Create datasets Import datasets Share datasets Version datasets Transform datasets Speed up data loading

Data connections

S3 buckets GCS buckets R2 buckets

Cloud folders Artifacts Switch clouds Use LitData

Data Prep Examples

Create a spatial dataset Embed wikipedia for $5 Load parquet faster Prepare 1T tokens Process 400M images Streaming benchmarks Tokenize 2M articles

Security

Data processing addendum (DPA)HIPAA SOC 2 Type II Self-serve due diligence Subprocessors

Deployment options Data collection Data retention Data residency Deleting my data

Security features

Audit logs Security Multi-factor authentication Custom encryption keys Disk encryption Configure SSO Secrets RBAC

Trusted platform

Internal access Secure software development Penetration testing Data backup

Team Management

Manage members Team management Manage cloud accounts Manage teamspaces Discoverability Manage costs Share private Studios Role-based access control

Manage members Drive Manage costs Secrets

Developers

Studio Batch jobs Multi-machine jobs Clusters Deployments Virtual Machines

Python Sandbox SDK JS Sandbox SDK

Studios Batch jobs Multi-machine jobs Sandbox Teamspace Drive

Docs/Lightning AI/Inference/Lightning Inference

Lightning Inference

Lightning Inference lets you deploy autoscaling, high-performance inference APIs.

There are three main ways to deploy on Lightning AI:

Deploy a container
Deploy a Studio
Deploy a turn-key model

Was this page helpful?

Thanks for the feedback.

What went wrong?

Inaccurate or out of dateHard to understandMissing informationHard to find things / poorly organizedCode sample didn't workSomething is broken (link, image, layout)

Share with AI

Open in ChatGPT Open in Claude

Was this helpful?

Thanks for the feedback.

What went wrong?

Inaccurate or out of dateHard to understandMissing informationHard to find things / poorly organizedCode sample didn't workSomething is broken (link, image, layout)

↑↓ Navigate↵ Open⌘↵ New tabesc Close