Phi, Gemma, TinyLlama — small language models that run on devices, enabling offline AI and privacy-first applications.
Five passes over the same idea, each from a different angle. Do them in order, or jump to whichever you need.
Small language models (1-7B parameters) run on consumer hardware, phones, and edge devices. Models like Phi-3, Gemma 2, and TinyLlama offer surprising quality for their size. Key techniques: distillation, quantization (GGUF Q4), ONNX runtime, and WebLLM for browser inference.
