aicost

A simple, efficient library for calculating AI model costs across various providers. It's completely type safe.

Installation

npm install aicost

Usage

Calculate Cost

Calculate the cost of using an AI model based on input and output amounts.

import { calculateCost } from 'aicost'

const cost = calculateCost({
    provider: 'openai',
    model: 'gpt-3.5-turbo',
    inputAmount: 6032,
    outputAmount: 1238
})

console.log(cost)

{
  inputCost: 0.18096,
  outputCost: 0.07428,
  inputCostUnit: "token",
  outputCostUnit: "token",
}

For Google generation responses, pass the complete usageMetadata object. This lets aicost include thinking tokens, tool-use prompts, cached input, and model-specific long-context rates:

const cost = await calculateCost({
    provider: 'vertex-ai',
    model: 'gemini-2.5-pro',
    inputAmount: response.usageMetadata.promptTokenCount,
    outputAmount: response.usageMetadata.candidatesTokenCount,
    usageMetadata: response.usageMetadata,
})

When usageMetadata is present, its prompt, tool-use prompt, cached-content, candidate, thinking, and total counts are authoritative. The reported total prevents tool-use breakdowns from being counted twice across Google API variants. The effective prompt size selects the applicable context tier; Google charges all input and output tokens at the long-context rate when the prompt exceeds 200,000 tokens.

Get information for a model

Retrieve detailed information about a specific AI model.

import { getModelInfo } from 'aicost'

const modelInfo = getModelInfo({
    provider: 'openai',
    model: 'gpt-4'
})

console.log(modelInfo)

List supported providers

List all available AI model providers.

import { getProviderList } from 'aicost'

const providers = getProviderList()

console.log(providers.includes('anthropic'))

List supported models from a provider

Get a list of all models offered by a specific provider.

import { getModelList } from 'aicost'

const models = getModelList('cohere')

console.log(models)

Credits

The information present on this package is extracted from the amazing work done at LiteLLM ↗, if you're using python, check them out!