Precios y ahorro de prompt caching

Cuando muchas solicitudes empiezan con el mismo prefijo largo (prompt de sistema, definiciones de herramientas, un documento compartido), el proveedor puede cachearlo y cobrar esos tokens a una tarifa mucho menor. Chatbots y agentes ganan más porque reenvían el mismo contexto en cada turno.

Calcula tu factura mensual con este descuento →

Modelo	Entrada normal /1M	Entrada en caché /1M	Ahorro
MiMo V2.6 Pro	$0.435	$0.0036	99%
MiMo V2.6 Flash	$0.14	$0.0028	98%
DeepSeek V4 Flash	$0.30	$0.006	98%
Claude Fable 5.1	$10.00	$0.25	98%
Claude Mythos 5.1	$10.00	$0.25	98%
DeepSeek V4 Pro	$1.32	$0.044	97%
GPT-6.1 Sol	$2.00	$0.10	95%
Claude Sonnet 5.5	$2.00	$0.10	95%
Claude Opus 5.5	$4.00	$0.20	95%
Kimi K3	$3.00	$0.30	90%
Gemini 3.5 Flash	$1.50	$0.15	90%
Mistral Medium 3.5	$1.50	$0.15	90%
Claude Sonnet 4.6	$3.00	$0.30	90%
GPT-5.3 Codex	$1.75	$0.175	90%
GPT-5.2	$1.75	$0.175	90%
Claude Sonnet 4.5	$3.00	$0.30	90%
Magistral Medium (latest)	$1.50	$0.15	90%
Mistral Medium 3	$1.50	$0.15	90%
Claude Haiku 5.5	$0.10	$0.01	90%
GPT-6 Luna	$0.10	$0.01	90%
Mercury 2.5	$0.20	$0.02	90%
Gemini 3.8 Flash	$0.75	$0.075	90%
Gemini 3.7 Flash	$0.75	$0.075	90%
Gemini Flash (latest)	$0.75	$0.075	90%
Gemini 3.6 Flash	$0.75	$0.075	90%
GPT-5.6 Luna	$0.20	$0.02	90%
GPT-5.4 mini	$0.75	$0.075	90%
GPT-5.4 nano	$0.20	$0.02	90%
Devstral Medium (latest)	$0.40	$0.04	90%
GPT-5 nano	$0.05	$0.005	90%
Gemini 2.5 Flash Lite	$0.10	$0.01	90%
Ministral 3b (latest)	$0.10	$0.01	90%
Devstral Small (latest)	$0.10	$0.01	90%
Ministral 14b (latest)	$0.20	$0.02	90%
GPT-6 Sol	$2.00	$0.20	90%
GPT-6 Astra	$10.00	$1.00	90%
Claude Opus 5	$5.00	$0.50	90%
GPT-5.6 Sol	$4.00	$0.40	90%
GPT-5.6 Terra	$2.00	$0.20	90%
Claude Sonnet 5	$2.00	$0.20	90%
Claude Mythos 5	$10.00	$1.00	90%
Claude Fable 5	$10.00	$1.00	90%
Claude Opus 4.8	$5.00	$0.50	90%
GPT-5.5	$5.00	$0.50	90%
Claude Opus 4.7	$5.00	$0.50	90%
GPT-5.4	$2.50	$0.25	90%
Gemini 3.1 Pro Preview	$2.00	$0.20	90%
Claude Opus 4.6	$5.00	$0.50	90%
Claude Opus 4.5	$5.00	$0.50	90%
GPT-5.1	$1.25	$0.125	90%
GPT-5	$1.25	$0.125	90%
Gemini 2.5 Pro	$1.25	$0.125	90%
Gemini Pro (latest)	$2.00	$0.20	90%
Mistral Large 4	$0.68	$0.068	90%
Gemini 3.5 Flash Lite	$0.30	$0.03	90%
Gemini Flash Lite (latest)	$0.30	$0.03	90%
Gemini 3.1 Flash Lite	$0.25	$0.025	90%
Mistral Small (latest)	$0.15	$0.015	90%
MiniMax M2.5 Lightning	$0.30	$0.03	90%
MiniMax M2.5	$0.30	$0.03	90%
MiniMax M2.1	$0.30	$0.03	90%
MiniMax M2.1 Lightning	$0.30	$0.03	90%
Gemini 3 Flash Preview	$0.50	$0.05	90%
MiniMax M2	$0.30	$0.03	90%
Claude Haiku 4.5	$1.00	$0.10	90%
GPT-5 mini	$0.25	$0.025	90%
Gemini 2.5 Flash	$0.30	$0.03	90%
Mistral Large (latest)	$0.50	$0.05	90%
Ministral 8b (latest)	$0.15	$0.015	90%
Codestral (latest)	$0.30	$0.03	90%
Magistral Small (latest)	$0.15	$0.015	90%
Qwen3.8 Flash	$0.15	$0.016	89%
Muse Spark 1.3	$1.25	$0.15	88%
Muse Spark 1.2	$1.25	$0.15	88%
Muse Spark 1.1	$1.25	$0.15	88%
Qwen3.8 Max	$2.00	$0.25	88%
Grok 4.5	$2.00	$0.30	85%
Grok 4.3	$1.25	$0.20	84%
Grok 4.20 Multi Agent	$1.25	$0.20	84%
Grok 4.20 Non Reasoning	$1.25	$0.20	84%
Grok 4.20 Reasoning	$1.25	$0.20	84%
Kimi K2.5	$0.60	$0.10	83%
Kimi K2.6	$0.95	$0.16	83%
GLM-4.7	$0.60	$0.11	82%
GLM-4.6	$0.60	$0.11	82%
GLM-5.3	$1.40	$0.26	81%
GLM-5.2	$1.40	$0.26	81%
GLM-5.1	$1.40	$0.26	81%
Doubao Seed 2.1 Pro 260628	$0.8625	$0.1725	80%
Doubao Seed 2.1 Turbo 260628	$0.43125	$0.08625	80%
Qwen3.7 Plus	$0.40	$0.08	80%
Qwen3.7 Max	$2.50	$0.50	80%
GLM-5.3 Flash	$0.15	$0.03	80%
Kimi K2.7 Code	$0.95	$0.19	80%
MiniMax M3	$0.30	$0.06	80%
Grok Build 0.1	$1.00	$0.20	80%
GLM-5	$1.00	$0.20	80%
Grok Code Fast 1	$1.00	$0.20	80%
Grok 4.7	$2.00	$0.50	75%
Grok 4.6	$2.00	$0.50	75%
o3	$2.00	$0.50	75%
GPT-4.1	$2.00	$0.50	75%
GPT-4.1 mini	$0.40	$0.10	75%
GLM-5 Code	$1.20	$0.30	75%
GPT-4o	$2.50	$1.25	50%
GPT-4o mini	$0.15	$0.075	50%

Preguntas frecuentes

¿Escribir en la caché cuesta más?

Algunos modelos, sobre todo de Anthropic, cobran un recargo la primera vez que se cachea un prefijo (escritura en caché en las páginas de modelo). Los modelos sin tarifa de cache-write publicada cachean automáticamente y la calculadora cobra esas escrituras al precio normal de entrada.

¿Cuánto dura un prompt en caché?

Normalmente minutos (a menudo unos 5) salvo que nuevas solicitudes lo renueven; algunos proveedores venden cachés más largas.

¿Qué tasa de aciertos es realista?

Los bucles de agentes y chats de varios turnos suelen llegar al 50–80% de la entrada; las solicitudes únicas con documentos distintos rondan el 0%.