Why LLM price comparisons need source and channel tracking, not just rates

Comparing LLM API prices becomes unreliable when a single per-million-token column hides differences between providers and routing platforms. The author argues each price should be tied to its model, sales channel, token type, and source snapshot. Direct provider and router listings for the same model are kept as separate records, so a price gap is not mistaken for a provider rate change. Cost estimates are built from explicit workload inputs such as fresh input, cached input, output tokens, and monthly requests. Unknown rates are flagged as pending verification or defaulted to the full input rate rather than treated as zero, and the author describes using this approach in a source-linked pricing comparison tool.