Cross‑platform inference engine for huge AI models (1B–397B). Runs on any CPU (x86_64/ARM64) with AVX2/NEON, supports dense & MoE models (Qwen, Llama, Mistral…). GPU backends (Metal, OpenCL, CUDA) coming soon. No Python, no frameworks – pure C with optional PyQt5 GUI.
metal neon opencl x86-64 cuda moe avx2 arm64 pyqt5-desktop-application tui-app apple-silicon qwen ai-local cpu-reference
-
Updated
Jun 2, 2026 - C