aiapicostcalcUnit economicsPricesCompareGuidesChanges

Prompt caching cost savings

Guide · sources linked inline · 2026-10-26

When many requests share the same prefix (system prompt, tool definitions, a long document), providers can bill the repeated part at a lower cached-input rate.

Cached vs normal input price

ModelInput /1MCached /1MDiscount
GPT-6 Astra$10$190%
GPT-6.1 Sol$2$0.195%
GPT-6 Sol$2$0.290%
GPT-6 Luna$0.1$0.0190%
GPT-5.6 Sol$4$0.490%
GPT-5.6 Terra$2$0.290%
GPT-5.6 Luna$0.2$0.0290%
GPT-5.5$5$0.590%
GPT-5.4$2.5$0.2590%
GPT-5.4 mini$0.75$0.07590%
GPT-5.4 nano$0.2$0.0290%
GPT-5 mini$0.25$0.02590%
GPT-4.1$2$0.575%
GPT-4o mini$0.15$0.07550%
o4-mini$1.1$0.27575%
Claude Fable 5.1$10$0.2598%
Claude Opus 5.5$4$0.295%
Claude Opus 5$5$0.590%
Claude Sonnet 5.5$2$0.195%
Claude Sonnet 5$2$0.290%
Claude Sonnet 4.6$3$0.390%
Claude Haiku 5.5$0.1$0.0190%
Claude Haiku 4.5$1$0.190%
Gemini 4 Argon$2$0.195%
Gemini 3.8 Flash$0.75$0.07590%
Gemini 3.6 Flash$0.75$0.07590%
Gemini 3.5 Flash-Lite$0.3$0.0390%
Gemini 3.1 Pro Preview$2$0.290%
Gemini 3.1 Flash-Lite$0.25$0.02590%
Gemini 2.5 Pro$1.25$0.12590%
Gemini 2.5 Flash$0.3$0.0390%
Gemini 2.5 Flash-Lite$0.1$0.0190%
Grok 4.7$2$0.575%
Grok 4.3$1.25$0.284%
Grok Build 0.1$1$0.280%

Things to know

Next step → Try caching % in the calculator

Prices last checked 2026-10-26 against official sources · how we verify

Price-change alerts

One short email when a provider changes a price or a promo ends. No spam, unsubscribe any time.