CUDA Oxide 0.3 Vastly Improves Rust Support For NVIDIA CUDA Programming

NVIDIA software engineers released CUDA Oxide 0.3 for closing out October. CUDA Oxide as a reminder is their experimental Rust to CUDA compiler that has been seeing a lot of activity in recent months. With CUDA Oxide 0.3, cuda-oxide is now using the official stable Rust version, currently Rust 1.99.1. In prior versions of cuda-oxide, running Rust nightly builds were required. The new release also unifies the host runtime with cutile-rs in the cuda-rust monorepo. Another big step forward for cuda-oxide 0.3 is shipping an experimental version of "cute-rs" as the CUTLASS CuTe in Rust. With cute-rs it's able to run end-to-end Qwen3-8B at BF16 + NVFP4 along with Qwen3-TTS. Those models with cute-rs are being tested on NVIDIA Blackwell GPUs. For improving the debugging experience, v0.3 also has full cuda-gdb DWARF debugging support. Plus there is now in-kernel event tracing, register and spill diagnostics, and improved testing. The new release also brings performance and compiler optimizations for Blackwell and Hopper GPUs. Downloads and more details for CUDA Oxide 0.3 via GitHub.