t0k
The first decentralized compute network where launching a token brings AI online. t0k turns token launches into live inference capacity. Every token is a node, and its trading routes into a shared pool that runs the models we serve — image, video, language, and audio — with inference metered in TOK and drawn from the pool. No subscriptions, no API bills. The more nodes launch and trade, the deeper the pool and the more compute the network can serve.
DEPLOY A NODE
A node deploys to a pump.fun bonding curve, minted from the t0k protocol wallet.
SWEEP CREATOR FEES
Trading accrues creator fees at the protocol fee account. A scheduled sweep claims them and forwards the balance to the reserve.
MINT TOK CAPACITY
Reserve depth denominates TOK, the unit of GPU time. Deeper reserve, more TOK the network can spend before refill.
SETTLE INFERENCE
Each request resolves a model, meters TOK, and settles its SOL cost from the reserve. Volume refills it; inference draws it down.
LAUNCH A TOKEN
Launch a token in seconds. From its first trade, its creator fees flow into the shared compute reserve.
LAUNCH ›RUN INFERENCE
Run frontier image, video, language, and audio models. TOK is metered from the reserve. No keys, no invoices.
BROWSE MODELS ›INSPECT THE NODES
Every token funding the reserve, its market depth, and how much it contributes, live.
VIEW NODES ›THE PROBLEM
Running AI costs money — GPUs aren't free, and someone has to pay for them. Subscriptions and API bills put that cost on the user. t0k moves it onto the market instead.
TOK — THE UNIT
TOK is the model-agnostic unit of GPU time. Each job draws a fixed amount from the reserve. Token rewards convert to TOK continuously, so the pool deepens as the platform is used and traded.
BUILD ON T0K
Point any app at the pool and run inference without managing GPUs or keys. Every call draws from the same reserve, so third-party usage strengthens TOK instead of competing with it.
READ THE DOCS ›