Skip to content

Search across models, tools, services, companies, providers, APIs, plans, articles and news.

Methodology

How data is collected, how metrics are computed, and where the limits are — transparent and complete.

Data sources

All analytics read only from AiDiscovery’s canonical database; analytics is never a second source of truth.

  • Entity catalog: companies, models, products, tools, providers, APIs and plans.
  • Append-only pricing history — each price change is a new record; previous records are never rewritten.
  • Tracked change events from the data ingestion pipeline.
  • Privacy-conscious search events — aggregates only; no individual data is stored.

Pricing methodology

Prices are aggregated within one unit and one currency only; token prices are never mixed with subscriptions.

  • Medians are USD per 1M tokens, from current token prices only.
  • Only USD prices are aggregated; other currencies are not mixed into the same metric.
  • Median instead of mean — robust against outliers.
  • Cost scenarios are labeled “Estimated” and are not provider invoices.

Entity counting

Counts come from real table rows with real creation dates. No historical values are fabricated for periods before tracking began.

Data freshness

Each data type has its own freshness window — pricing has a short one, historical articles a long one. Stale data is marked, never deleted.

Limitations

To read these figures correctly, keep these limits in mind:

  • “Tracked by AiDiscovery” does not mean “the entire AI market”.
  • Model or tool counts are not market share; share requires external adoption/revenue data that is not available yet.
  • No predictions are made; only what happened is described.
  • Missing data is never treated as zero — it is shown as “—”.