Model rates
Every model Deploy Forward prices, in USD per million tokens. Canonical rates are hand-verified against each vendor's own pricing page; every day an automated pass observes them against two independent public sources (LiteLLM and models.dev) and commits the result to the open repository — the full time series, with diffs, is public.
| Model | Input | Output | Cache read | Cache write | Observed |
|---|---|---|---|---|---|
| The live table loads from the open repository — view rates/observed.json on GitHub if it does not appear. | |||||
“Observed” states how the sources relate: verified twice means both independent sources match our canonical rate; review flagged means they don't, and a drift issue is already filed; vendor-verified only means neither aggregator lists the model and the rate stands on its vendor citation alone. Spend figures built from these rates are always estimates at public list prices — never your subscription bill. The canonical table ships in the open deploy-forward package.