You can compare the cost of running models in different venues, tiers, and regions. Enter the expected token usage and the table shows you the prices sorted by expected month cost.
List prices from each venue, updated hourly at data.kottos.ai/prices.json. A prompt over a model's long-context threshold is priced at that tier, marked long. A venue with no cache price charges cached tokens as input. Your provider bills you directly; the Kottos AI fee is not included.
What you pay depends on how much of each prompt is read from cache, and moving a conversation to another venue starts its cache cold. Kottos AI records every request you send, with no prompt text, and its Simulate tab replays your own traffic under another route so you see the cost before you switch.