Sources
We read the prices of AI model APIs from several sources every day, cross-check them against each other, and publish each price together with its reliability and the sources that back it.
| Source | Weight | Role | Licence | Last updated | Records | Status |
|---|---|---|---|---|---|---|
| Provider pricing page | 1 | authoritative | — | 10 | OK | |
| models.dev | 0.6 | corroborating | MIT | 25517 | OK | |
| LiteLLM | 0.55 | corroborating | MIT | 15745 | OK | |
| pydantic/genai-prices | 0.5 | corroborating | MIT | 0 | OK | |
| Hand-maintained fallback | 0.4 | corroborating | — | 8 | OK | |
| OpenRouter | 0.25 | corroborating | — | 1653 | OK |
How a price is set
Every source states a price independently.
The provider's own pricing page wins by rule, not by weight. When it is unavailable, the aggregated catalogues vote, weighted by how accurate each has proven for that provider.
When they agree, confidence rises. When they disagree, the disagreement is published rather than hidden, with the runner-up amount kept as a fallback.
The verified label is not a score; it means the provider's own page states the price and an independent source corroborates it.
Prices are a daily snapshot, not a live feed. The date each was last confirmed is published alongside it.