AI Model Pricing
464 models across 66 providers, normalised to cost per million tokens so the numbers are comparable. Sort by input, output or context window.
| Model | Provider | Input $/1M | Output $/1M | Your monthly cost | Context |
|---|
How to read this
Providers publish prices per token, which produces figures like $0.00000015 that cannot be compared by eye. Everything here is multiplied to a per-million-token basis, the unit people actually reason in.
Input and output are priced separately because the ratio varies widely. Some models charge four times more for output than input; others charge the same. For chat workloads output usually dominates the bill, so sorting by output price is often the more useful view.
What this does not model
Cache reads, batch discounts and provider rate limits all change effective cost. A model that looks expensive here can be cheaper in practice if your workload hits the prompt cache heavily. Use this to shortlist, then check the provider's own page before committing to a budget.
Where the data comes from
The catalogue is pulled from OpenRouter's public models API, which aggregates pricing across providers, and is refreshed daily. If a figure looks wrong it usually means the upstream catalogue has not yet caught up with a provider change.