Skip to main content

HLEAD

HLEAD is the ticker symbol of the model token for Model 25 ("Pipeline Win Predictor"), Hokusai's first reference model deployment on the Sepolia v2 stack. It is not a protocol-level token — it is a standard model token that happens to be the first one deployed under the v2 infrastructure.

What "the HLEAD stack" means​

Every Hokusai model token is deployed as a triple of three contracts. For HLEAD, these are:

ConstantContract typePurpose
HLEAD_TOKENHokusaiToken (ERC20)The tradeable token. Minted as DeltaOne rewards, bought/sold on the AMM. Controlled by the TokenManager.
HLEAD_PARAMSHokusaiParamsPer-model configuration: infrastructure fee split, oracle price, and other parameters used by UsageFeeRouter and InfrastructureCostOracle.
HLEAD_POOLHokusaiAMMThe bonding-curve AMM pool. Holds the USDC reserve. Routes buy and sell orders.

This triple pattern is the standard for every model token on Hokusai. Future models will each get their own ticker, token contract, params contract, and AMM pool — the same structure, different addresses. See Model Tokens & TokenManager for the architecture.

Sepolia addresses​

All HLEAD contracts are deployed on Sepolia testnet (chainId 11155111) only. No mainnet deployment exists.

ContractAddress
HLEAD token (HLEAD_TOKEN)0x9690580864274E57899a79bD97e8d7C6cAe0d7d5
HLEAD params (HLEAD_PARAMS)0xc7325cB1f179f404Bad7FE62B83B708181FAaD6d
HLEAD pool (HLEAD_POOL)0x726f46e15cb8F05F291C6337F497da9D5A2738ff
Model 25 in ModelRegistryModel ID 25 — verified via ModelRegistry

The canonical source of truth for these addresses is hokusai-token/deployments/sepolia-v2-latest.json. The Deployments page is the docs mirror.

The underlying model: Pipeline Win Predictor​

Model 25 predicts whether a B2B sales pipeline opportunity will close as a won deal.

FieldDescription
DomainB2B sales — binary classification on pipeline opportunities
Ground-truth fieldwon (boolean)
Optional enrichment fieldsactual_deal_value, close_date_actual
Model ID25

Data contributors submit labeled pipeline records. The model is re-evaluated after each accepted contribution batch, and DeltaOne rewards are distributed when verified performance improves.

See Supplying Data for how to contribute training data to Model 25.

Deployment parameters​

Deployed on 2026-04-28 via TokenDeploymentFactory. All values are sourced from sepolia-v2-latest.json.

Token parameters​

ParameterValue
Initial supply1,000 HLEAD
infrastructureAccrualBps8,000 (80% of API fees to InfrastructureReserve)
initialOraclePricePerThousandUsd$3.00 per 1,000 API calls

AMM pool parameters​

ParameterValue
CRR10% (crr = 100,000 ppm)
Trade fee30 bps (0.30%)
IBR duration7 days (604,800 seconds) — Launch Period
Flat-curve phase threshold$25,000 USDC reserves
Flat-curve price$0.01 per HLEAD
Initial reserve$100 USDC
NetworkSepolia (chainId 11155111)

During the flat-curve phase (while reserves are below $25,000), buys are priced at $0.01/token. Once reserves cross the threshold, the pool switches to CRR bonding-curve pricing (P = R / (CRR × S)).

Why HLEAD exists​

1. Validates the v2 stack end-to-end. HLEAD was the first deployment to exercise DeployableTokenManager, TokenDeploymentFactory, two-phase AMM pricing (flat-curve → CRR), InfrastructureCostOracle, UsageFeeRouter, and InfrastructureReserve working together on a live network.

2. Live integration target. Builders integrating against Hokusai can read on-chain state, simulate buys/sells, contribute training data, and call the inference API against a real model on Sepolia — before mainnet.

3. Frontend reference. The Hokusai site uses HLEAD_TOKEN, HLEAD_PARAMS, and HLEAD_POOL as hardcoded constants for the homepage model card, token-deployment UI examples, and integration tests. These constants are Sepolia-only; mainnet-deployed models are discovered dynamically through ModelRegistry rather than hardcoded.

How HLEAD relates to other model tokens​

HLEAD is structurally identical to every other Hokusai model token. Nothing about HLEAD is protocol-special or privileged.

  • HLEAD is not a governance token
  • HLEAD is not a protocol fee or utility token
  • HLEAD is not a network or gas token
  • HLEAD is not a stablecoin

When future models graduate, each gets its own ticker, token contract, params contract, and AMM pool — the same triple pattern, different addresses. HLEAD just happens to be the first deployed under v2. See Model Lifecycle for how models progress from training to token graduation.

How to use the HLEAD reference app​

Buy or sell HLEAD​

Use the HLEAD_POOL AMM address (0x726f46e15cb8F05F291C6337F497da9D5A2738ff) and testnet USDC (0x4fE61E343D9c7CB5C0D9DeE293F0Dbcf7C2Dd645).

Contribute training data​

Submit labeled pipeline opportunity records to Model 25. See Supplying Data for the data format, submission flow, and how DeltaOne rewards are calculated and minted.

Read on-chain state​

const { ethers } = require("ethers");

const provider = new ethers.JsonRpcProvider("https://rpc.sepolia.org");

const HLEAD_TOKEN = "0x9690580864274E57899a79bD97e8d7C6cAe0d7d5";
const HLEAD_POOL  = "0x726f46e15cb8F05F291C6337F497da9D5A2738ff";

const erc20Abi    = ["function totalSupply() view returns (uint256)"];
const ammAbi      = ["function getReserves() view returns (uint256 reserve, uint256 supply)"];

const token = new ethers.Contract(HLEAD_TOKEN, erc20Abi, provider);
const pool  = new ethers.Contract(HLEAD_POOL,  ammAbi,   provider);

const [supply, { reserve }] = await Promise.all([
  token.totalSupply(),
  pool.getReserves(),
]);

console.log("HLEAD supply:", ethers.formatUnits(supply, 18), "HLEAD");
console.log("Pool reserve:", ethers.formatUnits(reserve, 6),  "USDC");

Call the inference API​

Use Model 25 via the standard Using Models workflow. The inference endpoint accepts pipeline opportunity records and returns a won probability score.

Where these addresses come from​

Addresses are set at deploy time by TokenDeploymentFactory and recorded in hokusai-token/deployments/sepolia-v2-latest.json (canonical source). The Deployments page is the docs mirror. The Hokusai frontend hardcodes them as HLEAD_TOKEN, HLEAD_PARAMS, and HLEAD_POOL because HLEAD is a Sepolia testnet reference deployment; mainnet-deployed models are discovered dynamically through ModelRegistry.

Next steps​

Last updated on