— Quanteta LLM API Price Index

How we collect, process and publish the data. We publish only facts (official aggregated prices and specs) plus factual arithmetic computed from them — never editorial opinions, predictions or purchasing advice.

Data as of:

LLM API list prices change frequently (new models and price cuts are common) and vary by tier, region, batch / cache usage and time. These are list prices captured at the time shown; always verify the current price with the provider before relying on it.

What is on this site

Official aggregated per-token API list prices for large language models: input / output price ($/1M tokens), context window, max output, cache-read price and modalities. Free endpoints (both input and output $0) are excluded from the priced index. Every value is shown exactly as the source returned it.

Computed numbers (factual arithmetic only)

These are pure factual arithmetic over published prices — not recommendations or "value" verdicts. Because list prices change, the computed figures are as-of the capture time too.

Cross-checking (not a score)

Where a model can be matched across both sources, we surface whether the OpenRouter and LiteLLM input prices agree — a factual cross-check, not a score. We normalise model IDs by provider and match on name. Only where a match exists do we report whether the input prices agree (tolerance $0.01/1M) as "Cross-checked". Unmatched models show "—"; we never fill the gap by guessing.

Why we do not invent our own score

A composite "quality-per-price" score drifts toward opinion without a trustworthy cross-model benchmark, so we do not invent one. For capability we instead surface a third party's published score (Open LLM Leaderboard (Hugging Face)) as-is, with source attribution and an as-of date, and never say which model to use or call any model "best". Cheapest is not "best", and different benchmarks rank models differently.

Data Sources

Prices & specs
OpenRouter Models API · LiteLLM model prices dataset
Capability score
Open LLM Leaderboard (Hugging Face) (Apache-2.0) — as of 2026-05-25

Current coverage: 327 models, 51 providers.

About the capability data. Capability scores are the per-model public values from Open LLM Leaderboard (Hugging Face) (Apache-2.0, a source whose attributed redistribution is licensed), shown as-is. Our original primary candidate, LMArena (Chatbot Arena) Elo, was not used: the leaderboard's own scores carry no clearly-stated licence permitting redistribution, so per DG policy ("if unclear, do not use and disclose honestly") we declined it. Artificial Analysis's Intelligence Index was likewise declined (proprietary IP, reuse terms not confirmed). Because Open LLM Leaderboard (Hugging Face) evaluates open-weight models only and lags the newest releases, many models (including closed APIs) have no value and show "—"; we never fill the gap with a guess.

Update Frequency

Because LLM API prices change frequently, we collect via automated collectors. Pages are regenerated when the underlying data changes; the "Data as of" timestamp reflects the actual capture time and is never cosmetically refreshed.

Data Quality

Limitations & Caveats

About · Home