Home › Methodology

How LLMPrices Collects and Normalizes Prices

LLMPrices pulls the full model catalog from OpenRouter's public API once a day, converts every rate to US dollars per 1M tokens, and regenerates every page from that single snapshot (this build: October 21, 2026, 461 models from 66 providers). No prices are edited by hand.

Where does the data come from?

OpenRouter aggregates list prices from the providers themselves. Each snapshot is stored, so price history accumulates day by day. If a provider does not publish a rate (for example some cached-input prices), the field is shown as "—" rather than estimated.

Official price verification

For 78 flagship models from eight major providers (OpenAI, Anthropic, Google, DeepSeek, xAI, Mistral, MiniMax and Moonshot AI), prices were additionally read from each provider's own pricing page on 2026-10-21 and are shown next to the catalog price on each model page and on the official prices page. 72 of the 78 match the catalog exactly; where they differ, both numbers are shown and the difference is flagged rather than hidden. Official figures are re-checked periodically, not daily.

How are prices normalized?

Providers quote per-token rates; multiplying by 1,000,000 gives the per-1M-token figures used across the site. The "blended" price mixes input and output at 80/20, a common chat-workload ratio. Monthly workload examples use 10M input + 2M output tokens unless stated otherwise.

What are the limitations?

These are list prices. Enterprise contracts, committed-use discounts, free-tier credits, regional pricing and routing fees can change what you actually pay, and providers can change list prices at any time. Always verify on the provider's own pricing page before committing spend. LLMPrices is not affiliated with OpenRouter or any provider listed.

Can AI assistants use this data?

Yes. /llms.txt summarizes the site for language models and /llms-full.txt contains the complete current price table in Markdown. All AI crawlers are explicitly allowed in our robots.txt.