Tiered storage models in MongoDB can optimize latency and cost by segmenting data based on access patterns, using tag-aware sharding to automatically move documents between shards based on predefined criteria. This approach allows for efficient management of data that is frequently accessed and less frequently accessed, reducing the need for complex data migration processes.