This chapter motivates language models as a new form of intelligence and explains why wireless networks are central to their training, deployment, and application. It reviews the evolution from large language models to small language models, the shift from training-time scaling to test-time scaling, and the growing importance of deployment constraints. The chapter then analyzes latency, bandwidth, privacy, energy, and orchestration requirements in wireless systems and introduces the concept of ubiquitous intelligence, where reasoning and generation are distributed across cloud, edge, and device tiers. Finally, it summarizes how these requirements shape the organization of the book and the codesign between models and networks.