On-device AI is revolutionizing mobile privacy and offline capabilities. This question explores running Large Language Models and other AI models directly on smartphones and tablets.

Contributors can discuss frameworks like CoreML, TensorFlow Lite, and ONNX Runtime. Share optimization techniques, model quantization, and hardware acceleration.

What models run well on current devices? How does on-device AI compare to cloud-based solutions? What's the roadmap for 2026-2027?