Text this: Optimizing Latency and Energy Efficiency in Edge-Native Large Language Models (LLMs) for Autonomous Mobile Agents.