Skip to content

outbrain-inc/azure-openai-cost-estimator

Repository files navigation

Azure OpenAI Cost Estimator

A lightweight TypeScript utility for estimating the cost of Azure OpenAI API calls. This package fetches pricing data from the Azure Retail Prices API and provides accurate cost estimates for various Azure OpenAI models.

Features

  • Fetches up-to-date pricing data from the official Azure Retail Prices API
  • Supports all Azure OpenAI models (GPT-4, GPT-3.5 Turbo, o-family models, Embeddings, DALL-E, etc.)
  • Automatically handles current and future o-family models (o1, o3, o3-mini, o4-mini, etc.)
  • Caching of pricing data to minimize API requests
  • Customizable currency and cache TTL
  • Zero dependencies beyond Axios
  • Written in TypeScript with full type definitions

Installation

npm install azure-openai-cost-estimator

Usage

Basic Usage

import { AzureOpenAICostEstimator } from 'azure-openai-cost-estimator';

// Create an estimator instance
const estimator = new AzureOpenAICostEstimator();

// Estimate the cost of a GPT-4 call
async function example() {
  const costUsd = await estimator.estimateCost({
    region: 'eastus',
    model: 'gpt-4',
    promptTokens: 500,
    completionTokens: 1200
  });
  
  console.log(`Estimated cost: $${costUsd.toFixed(4)}`);
}

example();

With Custom Options

import { AzureOpenAICostEstimator } from 'azure-openai-cost-estimator';

// Create an estimator with custom options
const estimator = new AzureOpenAICostEstimator({
  currency: 'EUR',      // Use Euro instead of USD
  cacheTTL: 3600 * 1000 // Cache pricing data for 1 hour (instead of default 24h)
});

// Estimate the cost of an embedding model call
async function example() {
  const cost = await estimator.estimateCost({
    region: 'westeurope',
    model: 'text-embedding-ada-002',
    promptTokens: 1500
  });
  
  console.log(`Estimated cost: €${cost.toFixed(4)}`);
}

example();

Different Model Types

import { AzureOpenAICostEstimator } from 'azure-openai-cost-estimator';

const estimator = new AzureOpenAICostEstimator();

// Text completion model (GPT-3.5 Turbo)
async function estimateTextModel() {
  return await estimator.estimateCost({
    region: 'eastus',
    model: 'gpt-3.5-turbo',
    promptTokens: 500,
    completionTokens: 300
  });
}

// O-family models - works with all current and future o-models
async function estimateOFamilyModels() {
  // o3 model
  const o3Cost = await estimator.estimateCost({
    region: 'eastus',
    model: 'o3',  // also works with variant formats like 'o-3', 'O3', etc.
    promptTokens: 1000,
    completionTokens: 500
  });
  
  // o3-mini model
  const o3MiniCost = await estimator.estimateCost({
    region: 'eastus',
    model: 'o3-mini',  // also works with variant formats like 'o3mini', 'o-3-mini', etc.
    promptTokens: 2000,
    completionTokens: 1000
  });
  
  // Any future o-family models will work automatically
  // For example, when available:
  const o4Cost = await estimator.estimateCost({
    region: 'eastus',
    model: 'o4',
    promptTokens: 1000,
    completionTokens: 500
  });
  
  return { o3Cost, o3MiniCost, o4Cost };
}

// Embedding model
async function estimateEmbedding() {
  return await estimator.estimateCost({
    region: 'eastus',
    model: 'text-embedding-ada-002',
    promptTokens: 1000
  });
}

// Image generation model (DALL-E)
async function estimateImageGeneration() {
  return await estimator.estimateCost({
    region: 'eastus',
    model: 'dall-e-3',
    imageCount: 5
  });
}

API Reference

AzureOpenAICostEstimator

The main class for estimating Azure OpenAI costs.

Constructor

constructor(opts?: EstimatorOptions)

Options:

  • currency (optional): ISO currency code (default: 'USD')
  • cacheTTL (optional): Cache time-to-live in milliseconds (default: 24 hours)

Methods

estimateCost(input: CostInput): Promise<number>

Estimates the cost of an Azure OpenAI API call based on the provided parameters.

Parameters:

  • input: A CostInput object containing:
    • region: Azure region (e.g., 'eastus', 'westeurope')
    • model: Model name or family (e.g., 'gpt-4', 'gpt-3.5-turbo', 'text-embedding-ada-002')
    • promptTokens (optional): Number of prompt/input tokens
    • completionTokens (optional): Number of completion/output tokens
    • imageCount (optional): Number of images for image generation models

Returns:

  • A Promise that resolves to the estimated cost in the specified currency (default USD)

License

MIT

About

A lightweight utility to estimate the cost of Azure OpenAI API calls

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published