Cache Hits and Cache Pricing: Why Identical List Prices Produce Different AI Bills

Key Info

Requesty shared observed costs for two models with the same list price: Fable 5.1 came in at $1.87 per 1M tokens while Astra cost $2.90 per 1M tokens, with caching savings of 82% and 74% respectively.

Highlights

• Same list price does not mean the same bill: observed per-1M-token costs were $1.87 for Fable 5.1 and $2.90 for Astra.
• Caching reduced costs by 82% for Fable 5.1 and 74% for Astra.
• Repricing Astra's cache reads from $1/M to $0.25/M would cut its bill by 23%, showing cache pricing is as important as cache hits.