Releasing quantized versions of our Llama 1B and 3B on device models. Reduced model size, better memory efficiency and 3x faster for easier app development. 💪