What each AI model costs at your token mix

Input and output prices, context window and modalities for every model listing, priced per request and per month for the workload you describe.

Snapshot Loading the bundled snapshot… Free with a SkillSafe sign-in, no credits.

1. Your token mix

Input tokens per request Output tokens per request Cached share of input% Requests per month

2. Compare models

Search Maker Host Kind Context at least

Reads Has

3. Price map

Across

model price frontier: nothing else is cheaper with a larger window open weights

4. What changed

5. Add your own scores (optional)

Paste results from your own evals or a leaderboard you trust: one model per line, then a comma or tab and a number. The table gains a score column and a score-per-dollar column, and the price map can plot cost against score. Stays in this browser.

Open CSV

Questions

Where do the prices come from?

From models.dev, an open, MIT-licensed catalogue of AI models and the prices each API host lists for them. The page opens on a bundled snapshot; signed-in visitors can load the current catalogue through the app's server.

How is the cost per request worked out?

Input tokens times the input price, plus output tokens times the output price, with the cached share of the input billed at the host's cache-read price when it lists one. A prompt above 200,000 tokens uses the long-context price where the host lists one. Prices are USD per million tokens.

What is the price frontier?

The models for which no other model in the current view is both cheaper at your token mix and has a larger context window.

Does it cost anything?

No. The bundled snapshot works without an account. Loading live prices and saving snapshots needs a free SkillSafe sign-in and uses no credits.