Pay by card · OpenAI-compatible · See live prices →
OpenAI-compatible. Swap one line. Read how →

The cheapest inference
for every frontier model

A live market for AI tokens. Every request you send fills at the lowest live offer, with no change to your code.

No card to start · Pay with card, not crypto
Console · Live prices
● —
Best offer / 1M—
—
ModelOutvs list
Live book
Platform

Buy. Sell. Verify. Settle.

One market that matches unused API capacity with the apps that need it.

Buy

One endpoint, every frontier model. Each request fills at the lowest live offer with automatic failover.

Sell

List surplus credits at your own ask. When you're the best offer, traffic routes to you.

Verify

Every live offer gets an hourly test prompt. Offers that stop answering correctly leave routing until they pass again. It is a behavior check, not proof of identity: how checks work.

Settle

Per-request ledger. Buyers see the exact price of every fill; sellers get paid daily in USDC.

How it works

From your first request to the cheapest fill, in three steps.

01 · Connect

Change one line. Keep everything else.

Token Broker speaks the OpenAI Chat Completions, Responses and Anthropic Messages APIs. Point your SDK or coding agent at our base URL and use model ids from the live board; streaming and tool calls work as usual.

✓ OpenAI and Anthropic SDKs, Vercel AI SDK, LangChain
✓ Claude Code, Codex, OpenCode, Cursor
✓ Per-key budgets and rate limits
client.pypython
# base_url="https://api.openai.com/v1"
client = OpenAI(
  base_url="https://api.thetokenbroker.ai/v1",
  api_key="tb_…",
)
Models

Frontier models, market prices.

Best live offer per 1M tokens.

All models →
For sellers

Turn unused credits into revenue.

Team plans, committed spend, promo credits. List them at your price. Paid per fill in USDC, daily.

Requests routedlast 24h
—
fills recorded on the book

Ready to pay the market price?

Swap one line. Watch the bill drop.