AI API Cost Calculator
Estimate the real cost of any AI API call before you build. Pricing data is pulled live from models.dev, so GPT, Claude, Gemini, DeepSeek, Llama and thousands of other models stay up to date automatically. Enter token usage, factor in cache and reasoning tokens, project monthly spend, and compare up to 4 models at once.
Uso AI API Cost Calculator
Modelo
Uso
a peticiónEscenarios rápidos
Calculadora de tokens
Desglose de costes
| Componente | Fichas | Tasa / 1M | Coste |
|---|---|---|---|
| Selecciona un modelo e introduce el consumo para ver el desglose. | |||
Modelos para comparar
Uso compartido (por solicitud)
Comparación
| Modelo | Entrada / 1M | Salida / 1M | A petición | Mensual |
|---|
| Modelo | Proveedor | Contexto | Entrada / 1M | Salida / 1M | Lectura de caché / 1M | Estado | Uso |
|---|---|---|---|---|---|---|---|
| Escribe al menos un carácter para buscar en todo el catálogo. | |||||||
Leer el contenido
How to Use This Tool
Pick a provider and model from the live models.dev catalog, enter your token usage (input, output, cache, reasoning, audio), and the cost breakdown updates instantly. Use the request-volume slider to project monthly spend, or pick a quick scenario chip to jump straight in.
Features
- Live pricing - All rates are fetched from models.dev automatically and cached for 12 hours, so new models and price changes appear without manual updates.
- Full token breakdown - Input, output, cache read, cache write, reasoning, and audio tokens each billed at their own rate, exactly like the provider invoices you.
- Monthly projection - Set requests per month and see what your AI bill will look like.
- Model comparison - Compare up to 4 models side by side with the same usage profile, including a monthly-cost bar chart.
- Pricing browser - Search and sort the entire catalog by input, output, or cache-read price to find the cheapest model for your job.
- Token estimator - Paste text for a rough token count (~4 characters per token, English).
- Context warnings - The calculator flags when your input exceeds the context window or output exceeds the max-output limit.
Understanding the Numbers
Prices are shown in USD per 1 million tokens. When a provider does not publish a rate for a component (for example cache writes), the calculator falls back to the closest equivalent rate or hides that line item. Deprecated models are marked with a red badge so you can avoid pricing them into a new project.
Comentarios (0)
Describe el problema que has tenido para que podamos investigarlo y mejorar la herramienta.