About & methodology

Data source

All model specifications, capabilities and prices come from models.dev, an open-source, community-maintained database of AI models. An hourly job fetchesapi.jsonandmodels.json, validates them, and diffs the result against the previous snapshot.

How diffs work

• Every provider-model listing is flattened into comparable fields (costs, limits, capabilities, status).
• Field-level changes become typed events: repriced, deprecated, context_changed, capability_changed.
• A model appearing under a new provider emits provider_added; a brand-new canonical model emits model_added.
• Snapshots are committed to git, so history is auditable and reproducible.

Pricing caveats

• Prices are USD per 1M tokens as published by models.dev contributors.
• A dash (—) means no public price is listed — it is not necessarily free. Some providers genuinely list $0; we show those as $0 only when at least one side is non-zero or the provider documents a free tier.
• Tiered pricing (e.g., above/below 200K context) is flagged but shown at base rates.
• Always verify with the provider before purchasing.

For machines

Every view has a JSON twin — see /llms.txt for the index. Agents are first-class consumers here.