Build, tune, and deploy accelerated applications with CUDA

Find installation instructions, launch highlights, programming guides, compiler documentation, API references, CUDA libraries, profiling tools, samples, and release notes for the CUDA Toolkit.

Start with CUDA 13.3 highlights, then choose a path for installation, programming, compiler workflows, APIs, or libraries.

Install CUDA Release Notes Programming Guide

What's New in CUDA 13.3

Start with these CUDA 13.3 highlights, then open the release notes for the complete list of changes.

  • CUDA Tile C++ - Adds tile programming support in CUDA C++, with support in both NVCC and NVRTC.
  • CompileIQ - Uses AI-powered compiler auto-tuning to optimize NVIDIA GPU compiler settings for individual workloads and performance-critical CUDA kernels.
  • CUDA Python 1.0 - Introduces stable CUDA Python APIs for lower-level CUDA control and device-wide compute workflows.

Read the CUDA Toolkit Release Notes for the complete CUDA 13.3 feature list, resolved issues, known issues, deprecations, and compatibility notes.

Explore CUDA documentation

Find CUDA documentation by workflow, feature area, or API/reference type.

Release and installation

Programming

CUDA Compiler

PTX and low-level programming

CUDA APIs

CUDA libraries

Additional libraries

Architecture compatibility

Architecture tuning

Debugging and profiling

GPUDirect and platforms

Compiler SDK

Application notes and references