Gemma 4 launched under Apache 2.0 with four model sizes — but the official docs left gaps around mobile deployment, local setup, and model comparison. gemma4.app fills those gaps: → Live playground using Gemma 4 26B via OpenRouter, no signup needed → Mobile deployment guide for Android (LiteRT-LM, AI Edge Gallery) and iOS (MediaPipe SDK) — the two platforms have different stories → Local setup guides for Ollama, llama.cpp, LM Studio, and MLX