What a token is
A token is a chunk of text the model reads or writes — often about four characters or 0.75 English words, but code, JSON, and CJK break that rule. Input tokens are the prompt plus context (system text, retrieved documents, chat history, tool results). Output tokens are the completion. Providers bill those two streams at different rates. This guide does not convert words to tokens for you; prefer usage counters from API logs.
The $/M formula
Estimated USD per request is (input tokens / 1,000,000 × input price per million) + (output tokens / 1,000,000 × output price per million). That is the only v1 math TokenCost runs in the browser from a typed rate file. Example: 2,000 input and 500 output on a $2 / $12 model is $0.004 + $0.006 = $0.010 for that call.
Monthly = per-request × requests/day × 30
A working monthly sketch is per-request cost × requests per day × 30. It is not a calendar month (February and 31-day months differ slightly). Scale requests per day if you need a different window. The homepage calculator turns those three numbers into a sorted shortlist.
What list rates miss
Public list rates omit prompt-cache hits, batch or flex discounts, regional multipliers, thinking tokens billed as output, and non-text meters (images, audio, video, search grounding). Cached or overnight-batch traffic will come in under the list sketch. Multimodal apps should treat these numbers as the language slice only. Confirm on the vendor page before you lock budget.
Open the live calculator
The multi-provider table lives on the homepage so this URL can stay a guide. Open live calculator to paste tokens, filter, and share a link.