Cheapest 5 models per tier, sorted by completion price per million tokens. Frontier > $5/M, mid $0.5–$5/M, small < $0.5/M.
Largest 7-day completion price decreases across all tracked models. Updated nightly at 03:30 UTC.
Cheapest paid model for each common capability.
Models that currently report $0 for both prompt & completion tokens. Sorted by context window.
Counts, cheapest completion price, and free-model totals per provider on OpenRouter.
token-floor is an honest, live pricing index for AI inference. It watches the public OpenRouter model catalogue hourly and snapshots every model's per-token price into a local SQLite database. Snapshots accumulate, so we can show price history sparklines, week-over-week deltas, and a feed of biggest drops.
GET /api/v1/models,
unauthenticated, fetched every hour at :00.
GET /api/models/<slug>,
unauthenticated, fetched every 6 hours to enrich open-source
models with download count, likes, and pipeline tag.
Every number on this page came from a real network request. If a
fetch fails, the row simply doesn't appear that hour — the fetch
attempt is logged and visible at
/token-floor/api/fetch-log.
Tiers are derived from each model's completion price per million tokens:
JSON endpoints, CORS open, no auth.
GET /token-floor/api/statsGET /token-floor/api/models?provider=…&tier=…&feature=…&search=…&sort=…&order=…&limit=…&offset=…GET /token-floor/api/floorGET /token-floor/api/movers?days=7GET /token-floor/api/model/<id>GET /token-floor/api/providersGET /token-floor/api/fetch-logPOST /token-floor/api/refresh (rate-limited 1/min/IP)