A compact but capable text model that punches above its weight for its size class. The FP8 dynamic quantization means it runs efficiently on modest hardware while preserving most of the original Phi 4's reasoning ability — a practical trade-off between resource use and performance. It handles structured reasoning and instruction-following solidly, though its 16K context window is more modest than larger contemporaries.