How Hyperliquid works: HyperCore, HyperEVM and HyperBFT

Hyperliquid is a layer one blockchain whose core is an onchain exchange: the order books, the margin engine and every liquidation run inside the chain itself, not on a company server. Because of that, every order, every trade and every position that is still open can be read by anyone, which is what makes position-level data such as liquidation heatmaps and whale tracking possible.

This guide walks through the parts of the system in the order a trade touches them: consensus, the order book, the clearinghouse, the prices that drive margin, and the HyperEVM on the side.

One chain, two parts

Hyperliquid runs its own consensus algorithm, HyperBFT, which Hyperliquid describes as a variant of HotStuff consensus. Like most proof-of-stake chains, blocks are produced by validators in proportion to the HYPE staked to each of them.¹

The state of the chain is split into two parts that share this one consensus:

• HyperCore holds the exchange: perpetual and spot order books, the matching engine and the margin state of every address.
• The HyperEVM is a general-purpose smart contract platform in the style of Ethereum, which can read from and write to HyperCore.

The HyperEVM is not a separate chain or a rollup. It is secured by the same HyperBFT consensus as HyperCore, which is why a contract can interact with the order books directly, without a bridge in between.²

Hyperliquid's documentation states the performance targets in plain numbers: mainnet supports about 200k orders per second, and for a client located close to the validators the end-to-end latency of an order has a median of 0.2 seconds and a 99th percentile of 0.9 seconds.¹ These are the project's own figures; we do not measure them.

HyperCore: the order book

Each asset on HyperCore has its own order book. It works like the order book of a centralised exchange: prices are multiples of a tick size, sizes are multiples of a lot size, and orders are matched in price-time priority.³

Two details are specific to running the book onchain:

• Margin is checked twice. A new order is checked against the account's margin when it is placed, and the resting side is checked again at the moment it is matched. This keeps the margin system consistent even when prices move after an order was placed.
• The block orders actions by type. Within each batch of a block, actions that send no new resting or immediate orders come first, then cancels, then actions that send orders. Cancels therefore land before new orders in the same batch.³

HyperCore: the clearinghouse

The perps clearinghouse is the part of HyperCore that holds each address's balance and positions. Deposits are credited to the cross margin balance first, and positions open in cross margin by default. Isolated margin is also available: it assigns margin to one position, so its liquidation risk is separated from every other position of the account.⁴

Liquidations happen here too. When an account's equity falls below its maintenance margin, the clearinghouse first sends market orders to the book to close the position. Only if that does not work and equity falls further does a backstop take the position over (the liquidator vault, part of HLP).⁵ The liquidation calculator applies the documented formula to a position of your choice.

Oracle price and mark price

Two prices drive the system, and they are not the price of the last trade.

The oracle price is published by the validators for each perpetual roughly every three seconds. Each validator computes a weighted median of spot prices from several large exchanges and, for some assets, Hyperliquid's own spot market; the final oracle is the stake-weighted median of the validators' submissions.⁶ Funding is calculated against this price.

The mark price is the median of three inputs: the oracle price adjusted by a moving average of Hyperliquid's own premium, the median of best bid, best ask and last trade on Hyperliquid, and a weighted median of perpetual prices on other large venues. The mark is what margin, liquidations, trigger orders and unrealised profit and loss are measured against.⁷

The practical consequence: a position is liquidated when the mark reaches its liquidation price, not when a single trade prints there. A short wick on the book alone does not move the mark as far.

The HyperEVM

The HyperEVM brings Ethereum-style smart contracts to the same chain. A contract can read HyperCore order book prices through a read precompile and, once write system contracts are live on mainnet, send orders to the books. Hyperliquid describes the HyperEVM as an early stage of a gradual roll-out, with higher throughput and write contracts still to come on mainnet.²

Its blocks come in two sizes. Small blocks are produced every second with a 3M gas limit; large blocks every minute with a 30M gas limit, for bigger transactions such as contract deployments. An address chooses which kind it sends to.⁸

The public JSON-RPC endpoint supports the standard read methods but answers state queries only for the latest block, and it limits each IP address to 100 requests per minute. Historical state needs an archive node.⁹¹⁰

Why every position is public

On a centralised exchange, positions sit in a private database. On Hyperliquid they are part of the chain's state, and the public API returns them for any address. One documented request, clearinghouseState, takes nothing but an address and answers with that account's open positions, each with its size, entry price, leverage, margin used, unrealised profit and loss and its own liquidation price.¹¹

No key, login or permission is needed. That one fact changes what market data can show:

• Liquidation levels from real positions. A liquidation heatmap can add up the actual liquidation prices of open positions, instead of guessing them from assumed leverage. Our liquidation heatmap is built this way from the wallets we track; the guide How to read a liquidation heatmap explains the chart.
• Whale positions by wallet. Large positions can be followed address by address: opened, increased, reduced, closed or liquidated. Our whale tracker shows them with their liquidation prices.
• Liquidations confirmed by fills. A liquidated position leaves a fill that is marked as a liquidation, so a disappearing position can be checked against the wallet's own fills rather than inferred.

Our Hyperliquid exchange page shows the venue's flows and its protocol data (Assistance Fund, HYPE supply and HLP) in one place.

The limits are just as concrete. Reading every address all the time is not possible within the API's rate limits, so any tracker covers the wallets it polls, a few minutes apart. Our own pages say which wallets are covered and how old each position is.

Common questions

Is the HyperEVM a separate blockchain? No. HyperCore and the HyperEVM are two parts of the state of one chain, secured by the same HyperBFT consensus.²

Can anyone see my positions? Yes. Any address's open perpetual positions and margin summary can be read through the public info API by anyone who knows the address.¹¹

Which price triggers a liquidation? The mark price, not the last trade. The mark combines the oracle, Hyperliquid's own book and prices from other venues.⁷