Google DeepMind launched Nano Banana 2 Lite and Gemini Omni Flash to optimize edge-device performance. These models prioritize low latency and reduced memory footprints for real-time applications. Developers can now deploy high-speed multimodal capabilities on constrained hardware. This update targets efficiency over raw parameter count to enable faster on-device inference.