vla.cpp
A C++ inference engine built on llama.cpp for Vision-Language-Action robot policies, running open VLA models such as SmolVLA, GR00T and BitVLA as self-contained GGUF files without Python.
vla.cpp is a C++ inference engine, built on llama.cpp, for Vision-Language-Action models used to drive robots. It runs open VLA policies including SmolVLA, Pi-Zero, BitVLA, Evo-1 and GR00T N1.5 through 1.7 under one runtime, with each policy packaged as a single self-contained GGUF file that needs no Python or PyTorch at inference time.
The resulting binaries run on CPU, Apple Silicon, CUDA GPUs from consumer cards down to Jetson-class boards, Intel GPUs and NPUs through SYCL and OpenVINO, and Qualcomm Snapdragon CPUs, Adreno GPUs and Hexagon NPUs through OpenCL and a Hexagon backend. It is built with CMake, which fetches and pins the matching llama.cpp version automatically with no patch or submodule required.