Using LLMs to measure what LLMs cost and why smaller models aren't always cheaper

A discussion from reddit.com, titled "Using LLMs to measure what LLMs cost and why smaller models aren't always cheaper," highlights that a seemingly cheaper model per call can paradoxically increase overall workflow expenses. This occurs because while the model bill decreases, the cost of associated work, such as review time and corrections, rises. The issue often goes unnoticed as only the model bill is typically tracked, with later corrections rarely linked back to the originating task.