Last released Nov 19, 2026
TYLE (Tiled Inference Layout Encoding): a CPU-native single-file model format for local LLM inference.