Skip to main content

CUDA Playground · Workbench

Run CUDA, Python, or Rust on a real GPU

Write a kernel in the browser, run it on a managed NVIDIA GPU, and inspect the kernels, memory transfers, and timeline from that run. No local GPU or CUDA Toolkit.

  • CUDA C++ · Python (Numba) · Rust (beta)
  • Real NVIDIA GPU, every run
  • Free plan, no credit card
GPUFlight Workbench showing a Rust vector-addition project in the source editor with the run result beside it

Three languages, one profiler

Same GPU, same evidence, your language

Every project compiles and runs in the same managed sandbox, and every run produces the same kernel, transfer, and timeline evidence.

CUDA C++

Multi-file projects built with nvcc. The Hello, GPU and Vector addition templates start here.

__global__ void add(const float* a,
                    const float* b,
                    float* c, int n) {
  int i = blockIdx.x * blockDim.x
        + threadIdx.x;
  if (i < n) c[i] = a[i] + b[i];
}

Profiling target RTX 3090

Python with Numba

Launch and profile a CUDA kernel from Python, then verify the result on the host.

@cuda.jit
def add(a, b, c, n):
    i = cuda.grid(1)
    if i < n:
        c[i] = a[i] + b[i]

add[blocks, 256](a, b, c, n)

Profiling target RTX 3090

How it works

Edit, run, inspect

  1. 01
    Edit

    Start from a template or paste your own program. Each save is a revision.

  2. 02
    Run

    Compile and execute in a restricted GPU sandbox with a short time limit.

  3. 03
    Inspect

    Kernel launches, grid and block sizes, memory copies, NVTX ranges, and a zoomable timeline. Paid plans add metrics, SASS, and PC sampling.

GPUFlight Workbench result for a Rust run showing the kernel, memory transfers, and the GPU timeline

Choose your next CUDA experiment

Change a block size, an access stride, or an array layout. Understand what to measure before running the comparison.

Explore all CUDA performance guides

Common questions

CUDA Playground FAQ

Which languages can I write?

CUDA C++, Python with Numba, and Rust with cuda-oxide. Rust is in beta: its templates and toolchain can still change between releases.

Which GPU runs my project?

CUDA C++ and Python projects run on the RTX 3090. Rust projects use supported hardware such as the NVIDIA L4. Workbench shows the available profiling targets and any credit cost before you start a run.

Do I need an NVIDIA GPU or the CUDA Toolkit?

No. Compilation and execution happen in GPUFlight's managed environment. You only need a supported browser.

What does the free plan include?

500 credits / month for Performance Lab and Workbench, plus one Workbench project with Trace profiling. At most 5 hosted runs per day, shared across Performance Lab and Workbench. See the plan comparison for details.

Run a focused experiment

See what your code is doing on the GPU

Create a free project in CUDA C++, Python, or Rust and inspect the result.

Open the CUDA Playground

Want a guided problem with a target instead? Try Performance Lab.