How whattAI Sources and Labels Data
AI pricing, model capabilities, benchmarks, and vendor terms change quickly. This page explains what is live, what is curated, and what should be treated as an estimate.
Model Pricing
Model API pricing is fetched from OpenRouter when available and cached by Next.js for up to 12 hours. Prices are normalized to dollars per 1 million input and output tokens. If OpenRouter is unavailable, whattAI falls back to curated registry data from MongoDB.
Benchmarks and Quality Scores
Benchmark values are shown only when sourced from our curated registry (vendor publications, official model cards, or recognized leaderboards). whattAI does not fabricate or interpolate scores. Models without published benchmarks display a “Not published” state instead of a number, and are sorted below models with sourced benchmarks.
Freshness Signals
whattAI does not claim editorial review dates. Freshness is shown through two objective signals: a per source lastChecked timestamp written by each ingestion run, and a Git derived Updated date for content files reflecting the last commit that touched a page. The original Published date is preserved separately.
Tools and Agents
Tool and agent listings are curated registry entries stored in MongoDB. Pricing, pros, cons, use cases, and supported models should be checked against vendor websites before purchase, especially for subscriptions, enterprise plans, privacy terms, and usage limits.
Sponsored and Affiliate Links
whattAI may earn revenue from affiliate links, sponsored placements, display advertising, or partner offers. Sponsored placements should be labeled. Affiliate relationships should not override fit, pricing transparency, or user value.
Read Affiliate Disclosure