# API Price Index

> Live index of LLM API pricing, quality and value, computed from providers' published prices. Every figure is traceable to a source URL and a verification date — no hand-waved numbers. Built for developers choosing where to run a model.

Each ranking is recomputed from the underlying price and benchmark data on every rebuild. Prices are per 1M tokens unless stated. Quality is measured (LMArena Elo, Artificial Analysis Intelligence Index, MTEB), never asserted.

## Value & quality rankings
- [Best value LLM APIs (quality per $)](https://apipriceindex.com/best-value-llm-api): 52 models ranked by LMArena Elo ÷ blended price
- [Best intelligence per dollar](https://apipriceindex.com/best-intelligence-llm-api): 56 models by Artificial Analysis Intelligence Index ÷ price
- [Best speed per dollar](https://apipriceindex.com/best-speed-llm-api): 11 endpoints by first-party measured throughput (tokens/s) ÷ blended price
- [Best embedding model APIs (MTEB per $)](https://apipriceindex.com/best-embedding-model-api): 15 embedding models, 4 priced by MTEB ÷ dollar

## Cost guides (data-derived)
- [All LLM API cost guides](https://apipriceindex.com/guides): index of every guide, each computed live from the price index
- [Same model, different price](https://apipriceindex.com/guides/same-model-price-spread): median 68% price spread across hosts of the same open model, up to 1347%
- [Output tokens are the expensive ones](https://apipriceindex.com/guides/output-vs-input-pricing): output tokens cost a median 3.93x input tokens
- [Prompt caching: who offers it, what it saves](https://apipriceindex.com/guides/prompt-caching): 366 endpoints across 39 providers with cached-input pricing
- [Biggest context windows, and their price](https://apipriceindex.com/guides/context-window): median 256k, up to 1311k tokens per call
- [Truly open weights vs open-but-restricted](https://apipriceindex.com/guides/open-weight-licenses): 51% permissive, 20% API-only
- [Knowledge cutoff dates by model](https://apipriceindex.com/guides/knowledge-cutoff): median September 2025, freshest June 2026

## Machine-readable data (CC BY 4.0)
- [Full dataset (JSON)](https://apipriceindex.com/api/all.json): every tracked endpoint with verified price, source URL, verification date, specs — regenerated each rebuild
- [Price history (CSV)](https://apipriceindex.com/api/pricing-history.csv): one row per endpoint and date, a point only when the price actually changed
- [OpenAPI 3.1 spec](https://apipriceindex.com/openapi.json): machine description of all data endpoints, incl. per-model JSON and price badges
- [Pipeline health](https://apipriceindex.com/health.json): freshness of each data job — check before trusting the index
- [How to use this data](https://apipriceindex.com/use-this-data): license (CC BY 4.0), attribution, stability contract, MCP server install

## Notes
- Methodology: prices from provider pricing pages and the OpenRouter catalogue; benchmark scores from primary leaderboards (LMArena, Artificial Analysis, MTEB), each on a single dated snapshot to avoid cross-source noise.
- Every model page carries the per-provider price, its source link and the date it was verified.