⬢github Zig · 2.3K ★ +500 since we first saw it · pushed 35 min ago · Apache-2.0
ashhart/TensorFold
LLM Inference Engine for Metal, CUDA and Vulkan.
TensorFold is an LLM inference engine written in Zig that serves language models directly on Apple Silicon (Metal) and NVIDIA GPUs (CUDA), handling checkpoint loading, tokenization, and kernel execution. It ships as a self-contained binary requiring no Python or MLX installation, with a simple CLI to pull and serve models.
Why now: People are discussing it recently likely because of its unusual Zig-based, Python-free approach to local LLM serving and its notable agent-driven contribution workflow (ATC) for coordinating coding agents.
Who it is for: Developers who want to run LLMs locally on Macs or NVIDIA GPUs with a lightweight single-binary server.
aiai-toolsllmllm-inferencellm-tools
Stars over our 526 snapshots: 1.8K to 2.3K, since 5 d ago.
Where people talked about it
- ⬢github ashhart/TensorFold 2 d ago
API: https://socialmediatrends-api.osmike.com/v1/repos/ashhart/TensorFold