Coverage, not provider quality
AI API pricing evidence coverage ranking
Count fresh accepted price observations by provider without turning source coverage into a reliability, quality, or recommendation score.
| Coverage rank | Provider | Fresh observations | First-party evidence |
|---|---|---|---|
| 1 | WaveSpeedAI | 3 | |
| 2Tie | fal.ai | 2 | |
| 2Tie | Fireworks AI | 2 | |
| 2Tie | Google Vertex AI | 2 | |
| 2Tie | Kie.ai | 2 | |
| 2Tie | OpenAI API | 2 | |
| 2Tie | OpenRouter | 2 | |
| 2Tie | Together AI | 2 | |
| 9Tie | Replicate | 1 |
|
| 9Tie | RunPod | 1 |
|
Coverage is not endorsement: 19 fresh accepted observations counted across 10 providers. The table does not measure output quality, uptime, latency, support, limits, or total cost.
Count contract
What the ranking measures
Each row counts current, accepted price observations with a first-party source. Multiple observations can represent separate input and output meters or materially different configurations from the same provider.
Counts sort high to low. Equal counts use provider name, then provider slug. A full result set and at least two providers are required; otherwise the ranking is withheld.
Questions developers ask
Does more price evidence mean a provider is better?
No. The count measures APIDir's current source coverage, not provider quality, reliability, speed, support, model quality, or value.
Which observations count?
Only accepted observations labelled fresh whose fresh-until time has not passed. Quarantined, aging, stale, and expired observations are excluded.
How are equal counts ordered?
Equal observation counts share a coverage rank. Provider name and then provider slug determine only their stable display order. Paid placement never changes either value.
APIDir Delta
Get checked price changes, not noise.
One concise issue with source links, comparable configurations, and corrections.