Data Sources
Every provider pricing page and API documentation source we monitor. Prices are verified manually against these pages weekly. When something changes, it is recorded in the changelog. Found a stale price? Email us.
Prices are published per model on the pricing page. Batch pricing is 50% of standard.
Cache write/read multipliers differ by TTL (5 min vs 1 hour). Batch pricing is 50% of standard.
Google charges a storage cost per token-hour for cached content, separate from read cost. Prices differ above 200K context.
Meta does not sell direct API access. Prices are averaged across major hosting providers (Together AI, Fireworks, Groq). Individual provider prices vary.
Prices can vary significantly during peak hours. Listed prices are standard rates.
Spot and on-demand prices. Community Cloud vs Secure Cloud rates differ.
Marketplace pricing — varies by listing. Listed prices are representative averages.
See the methodology page for how these prices are used in calculations.