Pricing

Measured endpoints. Time-priced.

Every row is a published endpoint from the live catalog. The rate is what you pay while it runs; the output price is that rate divided by the throughput measured at the named workload.

Prices are served from the live catalog.

Read the machine-readable catalog at /catalog.json; the Worker renders this page from the same file.

How we compute this

Output dollars per million tokens = usdPerHour / (outputTokensPerSecond * 3600) * 1e6. The throughput is the measured output rate for the task named on that row, not a peak or a vendor claim.

Speedup appears only where the catalog publishes the baseline this endpoint was measured against. A missing measurement is shown as a dash, never inferred.