Calculator
LLM pricing comparison, at your volume.
Published rates are per million tokens, which tells you nothing until you multiply by your own traffic. Set your monthly volume and every tracked model reprices against it, at the cheapest provider actually serving it.
50M
10M
Cheapest option
$0.80
Ling-2.6-flash · Novita
Dearest tracked option
$13,500
o1-pro · OpenAI
Same-model overpay, worst case
$1,650
Identical weights, wrong host, per month
Showing the 60 cheapest of 371 priced models. Tap any row to see every provider serving that model. Read the cheapest API call report or browse the full catalog.
What this calculator does and does not assume
Rates are the published list prices we hold a verified record for, in USD per million tokens. No committed-spend discounts, no enterprise agreements, no free tiers. If you have negotiated rates, your real number is lower than what you see here.
Aggregator listings are excluded from provider-to-provider comparisons. A reseller price is a real way to buy a model, but it is not a company serving weights, so counting it would flatter the spread.
Cheapest per token is not cheapest per task. A model that needs a retry, or emits three times the output for the same answer, costs more than this table suggests. Benchmark scores sit next to every price in the model catalog, and the sourcing rules are in the methodology.
Stop estimating. Price your real traffic.
Compare reads your actual usage instead of a slider, and shows what the same calls would have cost on the cheapest verified host. Free, forever.