HIP-3 Deployers Compared: Every Hyperliquid Builder DEX

hyperliquid

Every HIP-3 explainer currently ranking on Google names one or two deployers and compares none of them. Several high-authority pages still list venues that have been fully delisted on chain since June. So here is the whole cohort, in one table, with the dead ones marked.

All figures below were computed on August 26, 2026 from the Hyperliquid info endpoint ({"type":"perpDexs"} and {"type":"metaAndAssetCtxs"} per dex) with fee revenue cross-checked against DefiLlama. The top of this table is seven days old and the bottom of it stopped moving in June, so re-pull before you rely on any single number.

The league table

Rank	Deployer	Ticker	Live markets	24h notional	Open interest	24h fees	All-time fees	Status
1	tradeXYZ	xyz	101	$2,185,674,334	$3,647,358,273	$299,359	$55,885,001	Live
2	EntropyIO	io	2	$66,194,725	$7,956,907	$5,982	$11,206	Live
3	Markets By Kinetiq	mkts	4	$15,558,505	$4,235,177	$5,754	$894,257	Live
4	Paragon	para	22	$3,915,880	$12,684,116	$717	$139,039	Live
5	HyENA	hyna	18	$2,055,282	$5,555,078	$1,064	$1,891,689	Live
n/a	Felix Exchange	flx	0	$0	$0	$0	$813,636	Delisted
n/a	Ventuals	vntl	0	$0	$0	$0	$255,991	Delisted
n/a	dreamcash Markets	cash	0	$0	$0	$0	$1,701,446	Delisted
n/a	Kinetiq legacy	km	0	$0	$0	n/a	n/a	Delisted
n/a	ABCDEx	abcd	0	$0	$0	n/a	n/a	Never traded

For scale, Hyperliquid's own core perps ran 232 markets, $7,502,763,223 of 24h notional and $9,452,407,595 of open interest in the same window. HIP-3 as a whole accounted for 23.30% of all Hyperliquid perp volume, so the builder layer is now roughly a quarter of the exchange.

Fees, and why they are nearly identical

Hyperliquid sets the HIP-3 fee schedule, not the deployer. The base is 2x the standard validator-operated perp rate, split evenly between Hyperliquid and the deployer. Deployers can configure an additional fee share upward, from 0 to 300%, but there is no setting that prices below the floor.

Two things fall out of this. First, "lower fees" is almost never a real differentiator between HIP-3 deployers, because they inherit the same schedule and the only large discount, growth mode, is a Hyperliquid switch rather than a deployer feature. Second, the implied rates for Paragon and Kinetiq are backed out of observed fee revenue against notional, not read off a published rate card, so treat them as estimates. HyENA is the only deployer in the group publishing a genuinely different number, and it is higher rather than lower.

Entropy's effective rate is the most precisely measurable of the group because the arithmetic closes: $66,194,725 of notional at 0.009% implies $5,958 of fees, and DefiLlama reported $5,982 for the same window, a ratio of 1.00. Entropy's own documentation states the 2x HIP-3 rate and the growth mode multiplier separately and never states the product. The full working is in Entropy fees explained, and the mechanics of the schedule itself are in HIP-3 explained.

The "second largest deployer" claim, measured properly

Entropy is currently described in several places as the second largest HIP-3 deployer. That is correct on one measure and wrong on another, and the difference is entirely about the length of the window.

The venue took its first trade on August 19, 2026. It has existed for six of the last thirty days, so a trailing-30d metric is dividing a week of activity by a month of window. That is not a criticism of the venue, it is a measurement artifact, and it resolves on its own by roughly late September if volume holds. But if you see "second largest HIP-3 deployer" attached to a 30 day chart, the chart does not say that.

The reverse caveat applies to the headline too. Second place here means about one thirty-third of trade.xyz's daily notional and roughly 0.2% of its open interest. trade.xyz holds approximately 99% of all HIP-3 open interest. Second is a real achievement on a two-market venue that is a week old, and it is still a very distant second. The direct head-to-head, including the one ticker both venues list, is in Entropy vs trade.xyz.

The five that are alive

tradeXYZ (xyz). The incumbent and effectively the category. 101 live markets covering crypto, US equities, indices, commodities and the pre-IPO cohort, including xyz:SPCX on SpaceX at $109.0M of 24h notional, xyz:UNITREE at $33.4M and xyz:ZHIPU at $7.2M. It runs growth mode, publishes a full tier table, and has no token. It also had the most instructive incident in the cohort: on July 27, 2026 the mark on its SK Hynix perp fell from $1,127.90 to $917.25 in seconds, and trade.xyz paid discretionary USDC compensation against a $1,115.50 reference price while stating that its oracle had worked as designed.

EntropyIO (io). Two markets, io:ANTH on Anthropic and io:SNDK on SanDisk, operated by Bursa Global Inc. Its frontend also aggregates 317 markets deployed by other parties, which is worth stating plainly because it means most of what you see on entropy.io is not Entropy's. Growth mode is enabled on both of its own markets. Background is in Entropy on Hyperliquid.

Markets By Kinetiq (mkts). Four live index markets including US500 and USTECH, with tokenized fee ownership via kmHYPE and a KNTQ token. Third by 24h fees despite a fraction of the market count, which tells you index products carry volume efficiently. Note the separate legacy km ticker in the dead half of the table.

Paragon (para). 22 live markets across equities, crypto indices and rates, with an explicit pre-listing pricing methodology published. The closest structural analogue to what Entropy is attempting, at a much smaller size, and with no token.

HyENA (hyna). 18 crypto perp markets collateralized in USDe, with a points program and an explicitly stated absence of a token. Its GOLD and SILVER markets are delisted, so despite the market count it is not competing in the real-world-asset lane. It is also the only deployer here publishing a fee above the base schedule.

The five that are dead, and the pages that still say otherwise

This is where most published HIP-3 content is wrong. These are on-chain confirmed rather than inferred: every asset under these tickers returns isDelisted: true, with $0 across notional, open interest and fees.

Ventuals (vntl). The one that matters most, because it ran precisely the product Entropy now runs. All fifteen markets delisted. Last non-zero fee day June 18, 2026. It cleared more than $650M of cumulative volume on pre-IPO perps for Anthropic, OpenAI and SpaceX, returned its 500,000 HYPE bond one to one, and never issued a token, despite a persistent set of farming guides that still describe how to qualify for one. The settled marks are preserved on chain: vntl:OPENAI at 1336.2, vntl:ANTHROPIC at 1619.3, vntl:SPACEX at 2109.5. Its own site title reads "Ventuals is sunsetting." If you arrived from one of those guides, the Ventuals alternative page is the correction.

Felix Exchange (flx). 16 markets delisted, last fee day June 20, 2026, $813,636 of all-time fees. One important distinction: the parent Felix protocol is a separate, active product. Felix Perps winding down is not Felix winding down, and several summaries have merged the two.

dreamcash Markets (cash). 17 markets delisted, $1,701,446 of all-time fees, the largest historical fee base of the dead group. The dreamcash interface itself still earns builder fees on Hyperliquid, so it appears alive in some dashboards while its HIP-3 dex is not.

Kinetiq legacy (km). Superseded by the mkts ticker. Its presence in older lists inflates the apparent deployer count.

ABCDEx (abcd). One registered asset that never traded. Zero candles, zero fees.

Meanwhile, as of this writing, the hiperwire HIP-3 explainer updated August 24, 2026 still lists Ventuals and Felix as live, Nansen's HIP-3 post from July still lists Felix, CoinGecko's May post puts trade.xyz at "over 90% of HIP-3 OI" when the figure is now 99.165%, and Bankless's piece dates to November 2025. None of them are careless sites. The category simply moves faster than an evergreen explainer can, which is why every number on this page is stamped.

What actually distinguishes one deployer from another

Since fees are close to fixed and the execution layer is identical, the real differences are narrower than the marketing suggests:

• What is listed. trade.xyz has the pre-IPO cohort, the commodities and the indices. Entropy has Anthropic and SanDisk. HyENA is crypto. This is the only difference that reliably matters, because you cannot trade a market that does not exist.
• The oracle methodology. Each deployer defines its own. Entropy's pre-IPO mark is a clipped five-minute EMA of its own order book with no external blend, which is a materially different risk profile from an oracle blending external venues. Detail in is Entropy safe.
• Open interest caps. A deployer-set ceiling that you will hit before you hit anything philosophical. io:ANTH sat at 66.6% of its $5M cap on August 26, 2026, raised from a lower cap the previous day.
• Behavior after a failure. The only two data points in the whole category, Ventuals on SpaceX and trade.xyz on SK Hynix, both ended in discretionary compensation. Neither was a protocol guarantee, and trade.xyz said in writing that its payout does not guarantee future reimbursement.
• Liquidity depth. On a thin market, your realized cost is the spread and the slippage, not the 0.009% fee. This is where trade.xyz's lead is most consequential.

How to re-pull these numbers yourself

Nothing here requires an API key.

• POST https://api.hyperliquid.xyz/info with {"type":"perpDexs"} returns the full deployer list including dead tickers.
• {"type":"meta","dex":"io"} returns a dex's assets, leverage caps, margin modes and its growthMode flag.
• {"type":"metaAndAssetCtxs","dex":"io"} adds marks, open interest and 24h notional per asset.
• DefiLlama's fee adapters cover most deployers and are the cross-check we use for the fee columns.

Our full sourcing rules, including what we will and will not publish without a primary source, are on the methodology page. If you want a place to start trading one of these markets, how to trade on Entropy covers deposits and order entry, and the airdrop page exists because half the traffic to this category is still looking for a token that does not exist.