TinyLlama v1.1 is the scrappy, resource-conscious member of the team — it runs on hardware that would make larger models laugh, fitting comfortably on edge devices and low-memory environments. With only 1.1B parameters and a 2048-token context window, it trades raw capability for extreme efficiency, making it a practical choice when compute is the bottleneck. Expect competent but modest language understanding; it won't match larger models in reasoning depth or nuanced generation.