▲ 281 today2k starsZig
Zig-powered LLM inference runs on Metal and CUDA without Python
What is TensorFold?
TensorFold is a command-line engine written in Zig that loads LLM checkpoints and generates text directly on Apple Silicon GPUs via Metal or on NVIDIA GPUs via CUDA. It performs tokenization and kernel execution itself, so you don’t need a Python runtime or ML libraries.
Who it is for
Developers who want to run LLMs locally on Apple Silicon or NVIDIA GPUs can skip Python and use a single binary.
Try it in 60 seconds
You need: A Mac with Apple Silicon or a Linux box with an NVIDIA GPU and the Zig compiler installed.
No one-line install. The README has the setup steps.
Apache-2.0 licenseUpdated todayZig646 open issues
New projects are not vetted. Skim the code, and run it in a fresh folder or a container before giving it your files or keys.
The story so far
Early Signal first spotted ashhart/TensorFold on November 12, 2026, when it had 1.9k GitHub stars. It has since grown to 2k. On its busiest day it gained 281 stars. It is published under the Apache-2.0 license and has 525 forks. Early Signal files it under Agents, Inference, Coding.
Questions about TensorFold
- Is TensorFold free and open source?
- Yes. ashhart/TensorFold is open source under the Apache-2.0 license, and the code is on GitHub at https://github.com/ashhart/TensorFold.
- What do I need to run TensorFold?
- A Mac with Apple Silicon or a Linux box with an NVIDIA GPU and the Zig compiler installed.