Tag
Google's Gemma 4 E2B is demonstrated running on an iPhone 17 Pro via MLX optimization, achieving ~40 tokens/second with 128K context and offline thinking mode for coding and math.