← The economics of intelligence

What can we learn from OpenRouter data?

The first question I have is simple: how many tokens are going through OpenRouter, and roughly how much are users spending on them?

Play the OpenRouter animation →

Tokens processed each day

Prompt and completion tokens combined.

Loading OpenRouter data…

How fast is token use growing?

—average monthly growth
—latest month compared with last year
Change from the previous complete month.

Estimated spending at today’s prices

This is not actual historical spending. The gray area runs from pricing every known token as input to pricing every known token as output. The dotted line assumes 80% input and 20% output.

Input-to-output price range80/20 estimate

Estimated cost per million priced tokens each day

The 80/20 estimate divided only by tokens for which we have a current price.

How fast is estimated spending growing?

Based on the 80/20 estimate at today’s prices. This is growth in the estimated retail value of inference, not confirmed OpenRouter revenue.

—average monthly growth
—latest month compared with last year
Change from the previous complete month.

What we cannot see: OpenRouter does not publish the split between input and output tokens, and prices change over time. Discontinued models and the combined “other” group also have no usable current price.

Source: OpenRouter daily rankings. Prices are the current prices returned by OpenRouter when this dataset was downloaded.