Skip to content
LaunchLaunch an agent

Cheaper AI from the inference flywheel on BSC

Your agent gets every AI model on one key. Credit it leaves unused and AI bought in bulk fill an inference pool on BNB Chain that keeps AI cheap, and $BINF holders pay up to 8% less for credit.

Launch an agentEstimate your budget

Example: Nova answers a question with Claude Sonnet 5 through the bInference Router, paid from its AI budget. Credit agents leave unused for 7 days moves to the pool with AI bought in bulk, and $BINF holders buy credit for less.

Example
AI models, one key
345
off AI for $BINF holders
Up to 8%
until unused credit moves to the pool
7 days
to launch an agent
0.005 BNB

See how much AI your agent gets

Enter a daily trading volume, then pick a tax and an AI model. We use today’s prices.

$
Tax on each trade
AI model
Answer size

15,000 words read, 750 written$0.060 an answer

Your agent’s AI budget

$1,323a week

7 days, oldest spent first$189 added a day

About 22,000 answers a week from Claude Sonnet 5, or one every 27 seconds.

Routed viabInference Router
Where the tax goes$2,100 a week from traders
AI budget 63%
$1,323
Treasury 27%
$567
Flap 10%
$210
Launch an agent

Today’s model prices, and the same volume every day.

How the flywheel turns

Follow one dollar your token earns. Flap keeps a tenth. Most of the rest becomes your agent’s AI and the remainder buys AI in bulk for the pool. Your token’s vault splits it the same way every time, and no one can change the split.

Of every dollar of tax, 63 cents go to your agent’s AI budget, 27 cents to the treasury and 10 cents to Flap, the launch platform. Your agent spends its budget on AI, and what it has not used after 7 days goes to the pool, together with the AI the treasury buys in bulk. $BINF holders buy credit for up to 8% less.

What your agent hasn’t used after 7 days
Flap keeps 10%. The vault splits the rest 70/30, set at launch.View the contract

Cheaper AI for $BINF holders

Credit that agents don’t use and AI the treasury buys in bulk keep the inference pool full. Hold $BINF, the bInference token, and the credit you buy costs up to 8% less. The discount starts when $BINF launches.

Buy credit

How the pool fills

Every trade keeps it turning

  1. Trades pay tax
  2. Agents get AI
  3. The pool fills
  4. Holders save

Up to8%off for holders

$BINF’s own trading tax goes into the pool too.

Your discount

The more $BINF you hold, the more you save

5M $BINF
1% off
6M $BINF
2% off
7M $BINF
3% off
8M $BINF
4% off
9M $BINF
5% off
10M $BINF
6% off
12.5M $BINF
7% off
15M $BINF
8% off

We count your lowest balance of the last 7 days. Credit you buy lasts at least 7 days.

Example: with 10,000,000 $BINF, Carol pays $94 for $100 of credit.

You pay for credit in BNB, as everyone does. $BINF only sets your discount.

Every model, one key

345 AI models from Anthropic, OpenAI, Google, DeepSeek and more. Agents pay from their AI budget, and $BINF holders pay up to 8% less than the usual price.

Agents pay the whole barHolders, the solid part

  1. Claude Opus 5.5Anthropic

    Agents pay $0.12

    Holders pay $0.11

  2. Kimi K3MoonshotAI

    Agents pay $0.090

    Holders pay $0.083

  3. Claude Sonnet 5Anthropic

    Agents pay $0.060

    Holders pay $0.055

  4. GPT-6 SolOpenAI

    Agents pay $0.060

    Holders pay $0.055

  5. Grok 4.7SpaceXAI

    Agents pay $0.055

    Holders pay $0.051

  6. Gemini 3.8 FlashGoogle

    Agents pay $0.023

    Holders pay $0.021

  7. DeepSeek V4.1 FlashDeepSeek

    Agents pay $0.0086

    Holders pay $0.0079

  8. GPT-6 LunaOpenAI

    Agents pay $0.0030

    Holders pay $0.0028

One answer reads about 15,000 words and writes about 750. Agents’ prices are what they pay, and holders’ are at the top discount. Every model’s price per token is at binference.io/api/v1/models.

See all 345 models, with usage and details

Connect your agent with two settings

Point your agent at our address and give it its key. Every call is paid from its AI budget. No card and no monthly bill.

Your agent calls binference.io/api/v1 with its own key. The bInference Router passes the call to the model it names, from Anthropic, OpenAI, Google and more.
Your agentClaude Code, Codex, OpenClaw, Hermes or your own codeAddressbinference.io/api/v1Keybinf_…AnthropicOpenAIGoogleand more
import OpenAI from "openai";const client = new OpenAI({});const reply = await client.chat.completions.create({  model: "anthropic/claude-sonnet-5",  messages: [{ role: "user", content: "Summarize today's trades" }],  max_tokens: 1024,});
  • OpenAI and Anthropic formats, streamed
  • Tools, images and web search as usual
  • Paid from its AI budget, balance any time
  • Prompts and answers are never stored

Plain answers

About fees, budgets and the $BINF discount.

Launch your agent’s token

It takes one screen and 0.005 BNB. Your agent gets its key as soon as the token is live.

Launch an agent