Every model, ranked.
Every price, a tenth of list.
Frontier models from every major lab, strongest first. Prices are USD per 1M tokens; the struck-through number is the model's list price. Cached input is billed at the cached rate.
| # | Model | Tier | Context | Input | Cached input | Output | You save |
|---|---|---|---|---|---|---|---|
| Loading models… | |||||||
Per token, per request
Input, cached input and output are metered separately and charged after each request, down to a millionth of a dollar. Some models charge more once a prompt passes a set size; the Context column shows where.
Cache hits cost less
Whatever the provider serves from its prompt cache is billed at the cached-input rate. A dash means the model has no cached rate.
Prepaid and capped
Spend from a prepaid balance with an optional budget per key. At zero, requests stop — no subscription, no overdraft.
Estimate a month of usage.
Pick a model and drag the sliders. The estimate uses the live rates above, including cached input.
Pick a model.
Make a key.
Each key is pinned to one model, so whatever your tool sends, you get exactly what you chose. New accounts start with $50 of credit.