Burn $CENT for cheaper AI

One API key for text and image models from every major lab, priced in cents. Credit you get by burning $CENT costs less than paying cash.

Get an API keySee what a cent buys

Right now a cent buys 1,000 Claude Sonnet 5.5 tokens at list price.

Click any lab's coin to feed it to cent, or

What a cent buys right now

List prices from OpenRouter, checked every 10 minutes. Credit from a burn buys more than this.

ModelTokens out per 1¢
500
1,000
1,000
20,000
2,667
1,667
8,333
833
667
1,136
about 7.7¢ an image
about 3.9¢ an image

Text and image models work on one key today. Voice, video and GPU minutes come next.

Smarts per cent

The cheapest model is rarely the best buy. Here every model on the key sits by what a cent buys against how it scores.

354045505560655001,0002,0005,00010,00020,000Output tokens one cent buysIntelligence indexClaude Opus 5.5: 500 output tokens per cent, intelligence index 57.6Opus 5.5Claude Sonnet 5.5: 1,000 output tokens per cent, intelligence index 56Sonnet 5.5GPT-6.1 Sol: 1,000 output tokens per cent, intelligence index 51.8GPT-6.1 SolGPT-6 Luna: 20,000 output tokens per cent, intelligence index 38.1GPT-6 LunaGemini 3.8 Flash: 2,667 output tokens per cent, intelligence index 40.9Gemini 3.8 FlashGrok 4.7: 1,667 output tokens per cent, intelligence index 46.4Grok 4.7DeepSeek V4.1 Flash: 8,333 output tokens per cent, intelligence index 39.5DeepSeek V4.1 FlashKimi K3: 667 output tokens per cent, intelligence index 43.6Kimi K3
Further right buys more per cent, higher scores better on the Artificial Analysis intelligence index as OpenRouter lists it. Models without a score are left out.

What would your month cost?

Set how you use models and see the bill on each one, worked out from the live prices above.

300
1,500
400

A month on Sonnet 5.5

$63.00

0.7¢ a call
9,000 calls

At list price. Credit from a burn costs less; the rate is set when $CENT launches.

The same month on every model

1,500 in and 400 out per call, 30 days.

Mixes

Several models chained into one call and billed once. Each note is worth what a typical run costs today.

8.5¢a typical run

Launch kit

Give it an idea for a coin. Get back a name, a ticker, a one-line pitch and a logo.

  1. Sonnet 5.5 writes the name, ticker and pitch0.8¢
  2. Gemini Flash Image draws the logo7.7¢
4.0¢a typical run

Thread and cover

Paste a long post. Get a three-part thread and a cover image to go with it.

  1. DeepSeek V4.1 Flash cuts it into a thread0.2¢
  2. Gemini Flash Lite Image makes the cover3.9¢
3.3¢a typical run

Second opinion

Ask two frontier models the same question, then have a third say where they disagree.

  1. GPT-6.1 Sol answers first1.0¢
  2. Opus 5.5 answers second2.0¢
  3. Gemini 3.8 Flash compares the two0.4¢

How burning works

You burnConnect a Solana wallet and choose how much $CENT to burn. cent shows the credit you get before you sign, and holds that quote for ten minutes.
cent credits youYour wallet burns the tokens. cent reads the burn from the chain and adds the credit to your balance, in cents.
You spendUse the key anywhere that takes an OpenAI-compatible base URL. Each call shows its price, and your balance drops by exactly that much.

Agents can burn and spend too

An agent with its own Solana wallet can burn $CENT, get a key and call any model on it. You set its budget, and every cent it spends shows up in your usage.

The full setup for agents is in llms.txt.

Questions

Is $CENT live?
Not yet. The contract address and the burn rate are published at launch. Until then the app runs on sample data.
Why is burned credit cheaper?
Credit bought with cash costs list price. Credit from a burn is priced below it, so the same balance buys more tokens.
What works with the key?
Anything that accepts an OpenAI-compatible base URL and key. That covers most SDKs, Cursor and the common agent frameworks.
What does a mix cost?
The sum of the model calls inside it, at the same per-token prices as single calls. You see it before it runs.