Skip to main content
Tulip logo

​
Open model credit markets

In version 1, Tulip gives each pinned Hugging Face model revision one canonical ERC-20 credit, one credit/USDG market, and an open marketplace of independent inference providers. Market creators launch the credit and its permanent Uniswap v4 pool. Providers compete to serve the exact registered model. Users can trade the credit or deposit it into an inference permit and spend it on model usage. Explore model markets Browse live model credits, provider availability, and current market status.
This documentation covers Tulip’s production environment on Robinhood Chain. Review every wallet transaction and verify on-chain activity in the Robinhood Chain explorer before signing.

Understand model markets

Learn how identity, credits, liquidity, providers, and settlement fit together.

Start using Tulip

Follow the user flow from Robinhood Chain to a paid inference request.

Launch a market

Create the canonical credit and one-sided credit/USDG pool for a pinned model revision.

Become a provider

Register an independently priced Hugging Face Inference Endpoint in version 1.

​
One market, many providers

Every model market contains:
  • One deterministic v1 ModelID for a Hugging Face repository, 40-character commit revision, and task.
  • One fixed-supply model credit.
  • One canonical Uniswap v4 credit/USDG pool.
  • One permanently locked launch position.
  • Any number of provider offers with independent tariffs and payout addresses.
The market creator does not need to operate inference hardware. A provider does not need permission from the creator to register an offer.

​
Two connected markets

Tulip separates price discovery from inference delivery.

Credit market

The native Uniswap v4 AMM trades model credits against USDG. Swap execution stays compatible with standard v4 routing and telemetry.

Inference market

Providers quote input, output, minimum, and wake charges in the model’s canonical credit. The gateway selects and meters an eligible offer.
This separation lets a credit develop an open market price while providers compete on inference cost and service quality.

​
Demand can arrive before supply

Tulip markets do not require an active inference provider at launch. A model credit can begin trading while the model is still unserved, allowing traders to signal demand before a provider commits capital to hosting it.
Until a provider comes online, the credit remains tradeable but cannot be used for inference. Once service begins, the same market becomes the payment layer for competing providers.

​
Where to go next