# about

ZML is a production inference stack that decouples AI workloads from proprietary hardware. From a single codebase it compiles large language models — Llama, Gemma, Qwen, Mistral — to run at peak speed across NVIDIA, AMD, Google TPU, Intel, and Apple Metal, letting teams mix accelerators, break vendor lock-in, and cut inference cost and energy. Its LLMD inference server, built on Zig, MLIR, and OpenXLA, launched free in 2026.

# project