One engine, any model. 100% Hugging Face model coverage & Zero python during runtime.
cpp vulkan mit-license quantization rocm inference-engine npu 1bit model-agnostic cplusplus-23 ai-inference llm-inference one-binary local-ai gguf open-source-ai amd-strix-halo xdna-2 zero-python 1bp-format
-
Updated
Aug 29, 2026 - MLIR