Model Choice, Memory, and Hardware

Leaderboards are discovery tools, not decisions. This module shows how to shortlist models for a defined task, reason about memory without memorizing a fake “minimum RAM” number, and match model size and context length to what the workload actually needs.