NVIDIA released CUDA Toolkit 13.4 on September 9, 2026, introducing native CUDA support for Windows on Arm. This update allows developers to compile and run CUDA applications on Windows Arm64 systems, closing a gap that existed since CUDA's inception in 2007. The toolkit offers two workflows: native compilation on Windows Arm64 machines and cross-compilation from x86-64 systems. Developers must install an R616-series Developer Driver to utilize the new features. The release aligns with the upcoming launch of NVIDIA's RTX Spark laptops, built on the N1X Arm platform. The toolkit is designed for RTX Spark devices, which leverage NVIDIA’s Blackwell GPU architecture. Current Windows on Arm devices can also begin using the toolkit in preparation for the hardware launch. The initial feature set for Windows Arm64 is narrower compared to x86-64, but developers are encouraged to adapt their applications for the new environment.