Pricing Methodology

This page explains how we collect, normalize, and present AI model pricing data on AI Pricing Guru.

Data Collection

Pricing data is collected daily from official provider pricing pages and API documentation. We scrape or manually verify pricing from each provider's public-facing pages:

Normalization Rules

All prices are normalized to USD per 1 million tokens for consistent comparison. Specific rules:

  • Currency: All prices are in USD.
  • Token unit: Always per 1,000,000 (1M) tokens. Providers that list per-1K prices are multiplied by 1,000.
  • Input vs Output: We track separate input and output prices.
  • Cached input: Where available, we track the cached/discounted input rate separately.
  • Batch pricing: Batch API discounts are noted but not used as the primary comparison price.

Model Status

Each model is tagged with a status:

  • Active: Currently recommended by the provider, generally available.
  • Legacy: Still available but superseded by newer models.
  • Preview: Available in preview/beta. Pricing may change.

Update Cadence

Our pricing data (/api/pricing.json) is updated daily. The lastUpdated field in the JSON reflects the most recent successful update. When a provider announces a price change, we aim to reflect it within 24 hours.

Limitations

  • We track publicly listed prices only. Enterprise or negotiated pricing is not reflected.
  • Some providers change pricing without public announcements. We catch these during daily checks but there may be brief delays.
  • Context window sizes and benchmark scores are sourced from provider documentation and may lag behind the latest releases.

Data Access

Our full pricing dataset is available as JSON at /api/pricing.json. This data is free for personal and editorial use. For commercial redistribution, please contact us at aust@aumedia.sk.

Last updated: April 2026