The pricing dataset, free to download and cite
Every rate on this site comes from one dataset covering 60 models across 17 providers, last verified on 2026-08-16. It is published here as JSON and CSV under CC BY 4.0: use it commercially, chart it, build on it, just credit LLM Cost Lab.
What is in each row
| Field | Meaning | Example |
|---|---|---|
model_id | Stable slug, also the model's URL on this site | gpt-5-6-sol |
model_name | Display name as the provider markets it | GPT-5.6 Sol |
provider | Company selling the API | OpenAI |
input_per_million_usd | USD per 1M input (prompt) tokens, standard non-cached rate | $5.00 |
output_per_million_usd | USD per 1M output (completion) tokens | $30.00 |
cached_input_per_million_usd | Discounted rate for cache-hit input tokens, empty where the provider does not offer one (27 of 60 models do) | $0.50 |
batch_discount_percent | Percentage off input and output when run through an async batch API, where published (18 models) | empty |
blended_per_million_usd | Single ranking figure at a 3:1 input-to-output token mix | derived |
last_verified | Date the snapshot was checked against provider pricing pages | 2026-08-16 |
notes | Caveats that change the real cost: surcharges, promo windows, currency conversions | free text |
Pulling it into code
The JSON endpoint sends Access-Control-Allow-Origin: *, so it works from a browser
as well as a server. There is no authentication and no rate limit; please cache it rather than
fetching it per page view.
# Python
import pandas as pd
df = pd.read_csv("https://llmcostlab.com/data/pricing.csv")
print(df.nsmallest(5, "blended_per_million_usd")[["model_name", "provider", "input_per_million_usd"]])
# JavaScript
const res = await fetch("https://llmcostlab.com/data/pricing.json");
const { models, lastVerified } = await res.json();
const cheapest = models.sort((a, b) => a.blendedPerMillionTokens - b.blendedPerMillionTokens)[0];How to cite it
If you publish a chart, a blog post, or a report built on this data, the citation below is what we ask for. A plain link is fine too.
LLM Cost Lab. "LLM API Pricing Dataset." Verified 2026-08-16. https://llmcostlab.com/data/
Questions about the dataset
Can I use this pricing data in my own project?
Yes. The dataset is published under CC BY 4.0, which means you can use it commercially, modify it, and redistribute it, as long as you credit LLM Cost Lab with a link back to https://llmcostlab.com. No API key, no rate limit, no signup.
What format is the data in?
Two formats. pricing.json is a single JSON object with a metadata header and a models array. pricing.csv is a flat table with one row per model, suitable for dropping straight into a spreadsheet or a pandas DataFrame.
How often does the dataset change?
It is refreshed whenever a tracked provider changes a rate or ships a model, and every model carries the date it was last checked. The current snapshot was verified on 2026-08-16. Both files are regenerated from the same source as the pages, so the download can never disagree with the site.
What is the blended rate column?
A single number for ranking models, assuming three input tokens for every output token, which is closer to real chat and retrieval traffic than either rate on its own. Ranking on input price alone flatters models with cheap input and expensive output.
Does the dataset include enterprise or committed-use discounts?
No. Every figure is the standard public list price for the provider's own API. Negotiated rates, committed-use agreements, reseller markups, and free tiers are all excluded, because none of them are public and none of them are comparable across providers.
Where this data shows up
The same numbers drive every page on the site. These are the places worth starting.