We're excited to announce RunMat Desktop: a fast local app for running high performance math computations.
MATLAB syntax, modern hardware, modern tooling
MATLAB-style code is still one of the fastest ways to write down engineering math. Arrays, plots, simulations, controls, signal processing, optimization: the syntax is compact because it matches how a lot of technical work is already written.
Your laptop probably has a serious GPU. Your desktop definitely does. Numerical code should be able to use that hardware without you having to learn how to program GPUs. You should be able to write math like:
X = linspace(0, 2*pi, 100_000);
Y = sin(X);
plot(X, Y);And run it super fast, if your hardware is up to the task.
RunMat is built around that idea.
RunMat takes MATLAB-syntax code and runs it through an open-source Rust runtime with a GPU acceleration engine that can fuse array work and keep tensors resident on device when that is the right move.
RunMat is designed to run math blazing fast.
What Desktop adds
RunMat Desktop is the quickest way to get started with RunMat. It uses the same runtime and gives you the pieces you expect when you are actually working:
- local, offline file system access
- fast, collaborative cloud projects when you want them
- workspace with editor, plots, variables, and run history
- high performance, native GPU plotting
- plot, code and variable history
- markdown notebook support with LaTeX, Mermaid diagrams and more
RunMat Desktop connects to your GPU through Metal, DirectX 12, or Vulkan, and provides a high performance, GPU accelerated platform for executing your math.
Situationally aware agent
RunMat Desktop also includes a project-aware agent.
It can inspect the files in your project, iteratively run MATLAB-syntax code through the same runtime, read errors, inspect variables, look at figures, and propose reviewable changes.
It's situationally aware of the project and the code you are writing. It can see plots, it can zoom in on them, it can see splices of matricies. It can propose changes to the code, and you can review and accept or reject them.
RunMat agent is designed to help you iterate on your math faster, and run more ambitious projects.
Local by default
RunMat Desktop runs the full runtime on your machine. Local projects use your local files. You can run scripts, inspect variables, and view plots offline.
Cloud projects and the agent require a network connection on the Pro and Team plans. Enterprise deployments for complete air-gapped, compliance-ready environments are available. Contact us for more information.
RunMat Desktop is available for macOS Apple Silicon, macOS Intel, Windows x86_64, and Linux x86_64.
Try it
If you like MATLAB syntax but want a modern, open, GPU-accelerated environment, try RunMat Desktop.
RunMat™ is a registered trademark of Dystr, Inc. MATLAB® is a registered trademark of The MathWorks, Inc. RunMat is not affiliated with, endorsed by, or sponsored by The MathWorks, Inc.Related posts

RunMat Runtime 0.5 is out: multi-file project support, and resolving MATLAB language semantics like Rust's compiler
RunMat Runtime 0.5 adds manifest-backed multi-file projects, named entrypoints, stronger MATLAB function/class/indexing semantics, and a shared compiler foundation for tooling, acceleration, and JIT work.

Inside the RunMat Runtime: a Rust-like compiler pipeline to resolve MATLAB language semantics
A technical look at RunMat's Rust-like compiler pipeline for resolving MATLAB semantics across projects, functions, classes, indexing, bytecode, tooling, acceleration, and JIT work.

Open-Source MATLAB Alternatives 2026: Speed, Compatibility & Ease of Use
Benchmarks, compatibility, and setup time for four free MATLAB alternatives. Which runs your .m files, which needs a rewrite, and which can you try in 5 seconds?
Enjoyed this post? Join the newsletter
Monthly updates on RunMat internals, development, and performance tips.
Download RunMat
Download RunMat for full performance, or use RunMat in your browser for zero setup.