Hyperliquid vs Lighter vs Aster vs GMX: Which Perp DEXs Actually Show You Whale Positions?
We called the public endpoints of 9 perp markets with no account or key, read the documents of Lighter, Aster and dYdX, and recorded what each shows on whale positions.
Published 2026-11-05 · Tested November 5, 2026 · Deep Blue Alpha
We called the public endpoints of 9 perpetual futures markets on November 5, 2026, without an account or API key, and recorded five things: a wallet’s positions, a public ranking, liquidation prices, funding and open interest. Hyperliquid and Pacifica scored 10 of 10. GMX scored 9. Jupiter scored 7. Variational and GRVT scored 4: they publish market totals, but we could read no wallet’s position. We did not test Lighter, Aster or dYdX, and report only what their documents state.
The size of a market says little about how much it shows. Variational ranks third by DefiLlama open interest, leaving aside tradeXYZ, and we could read no individual position there. Pacifica ranks far lower, at $57.1M, and shows nearly everything. Aster, second on the same measure, documents its position calls as signed and a “hidden position” feature that it says excludes those positions from public data.
Hyperliquid’s share depends on the measure: about 56% of open interest, 44% of 24-hour volume and 36% of 30-day volume on the same DefiLlama page on the same day. This score measures data availability only. It says nothing about safety, cost or whether a market is suitable for anyone.
What exactly did we test, and how?
Most perp DEX comparisons repeat fee tables and volume charts. We asked a narrower question that a reader can check: if you hold no account and no API key, what can you actually read about the large positions on this market?
For nine markets we read the developer documentation, then sent real requests from one connection between 12:40 and 12:53 UTC on November 5, 2026. Only read requests were sent. Nothing was signed, no account was created and no order was placed. We used no VPN or proxy, so where a market blocked our connection we recorded the block and stopped. Results may differ from another location.
Each market gets a score of 0, 1 or 2 on five checks, for a maximum of 10:
- Wallet positions: can you read one address’s open positions?
- Public ranking: can you find large accounts without already knowing an address?
- Liquidation price: is the level returned as a field for a specific position, so you need not model it?
- Funding: is the current rate readable by market?
- Open interest: is the total readable by market?
We did not send requests to Lighter, Aster or dYdX. Their published terms restrict US persons, and those of Lighter and Aster explicitly cover their APIs. They are not scored, and the section on them below reports only what their own documents state.
A 2 means we read the data live. A 1 means partial data, a limited form or a result we could not fully confirm. A 0 means the request was refused or we found no such data. “Not found” below does not mean a feature does not exist, only that we could not find or reach it. If your own account of a market differs, the method section lets you repeat the check.
Which perp DEXs show whale positions? The results
Public data test, no account or key. Tested November 5, 2026, 12:40-12:53 UTC. Ranked by data availability only, not by suitability for trading.
| Market | Wallet positions | Public ranking | Liq. price | Funding | Open interest | Score | Size: OI (DefiLlama) |
|---|---|---|---|---|---|---|---|
| Hyperliquid | Yes | Yes, 48,471 rows | Yes | Yes | Yes | 10 | $7.76B |
| Pacifica | Yes | Yes, 6,894 rows | Yes | Yes | Yes | 10 | $57.1M |
| GMX (Arbitrum) | Yes | Build by query | Yes | Yes | Yes | 9 | $237.8M (all networks) |
| Jupiter | Yes | Top 50 by market | Yes | Borrow rate | Not direct | 7 | $41.7M |
| Variational | Not offered | Not found | No | Yes | Yes, long and short | 4 | $920.9M |
| GRVT | No, 401 | Not determined | No | Yes | Yes | 4 | $429.4M |
| edgeX | No, key needed | Not found | No | Empty reply | Yes, total | 3 | $175.9M |
| Antarctic | Rejected | Not found | No | Rejected | Yes | 2 | $360.0M |
| QFEX | No, 401 | Not found | No | Not found | Yes | 2 | $175.0M |
Sizes: DefiLlama perps page, data clock November 5, 2026 11:45 UTC. Tests are ours. Lighter ($744.6M), Aster ($1.44B) and dYdX ($52.3M) were not tested and are covered from their documents below. tradeXYZ, listed by DefiLlama at about $4.46B, runs on Hyperliquid and is covered there.
The finding in one line: Variational, third by DefiLlama open interest leaving aside tradeXYZ, showed us no individual position, and one of the smallest markets we tested showed almost everything. Size and transparency are separate facts.
Which calls did we use for each market?
This table is here so anyone can repeat the test. We list the calls, not any address, and we did not publish any wallet we read. To find a test address, use the discovery route in the last column.
What we called and what came back
| Market | Position call | Result | How to find a large account |
|---|---|---|---|
| Hyperliquid | POST /info, clearinghouseState | Size, entry, leverage, profit, liquidationPx, funding paid | Public leaderboard file, with address and account value |
| Pacifica | GET /positions?account= | Side, size, entry, margin, liquidation_price | /leaderboard?limit=25000, with profit, equity and open interest |
| GMX | GET arbitrum.gmxapi.io/v1/positions?address= | Size, collateral, profit, liquidationPrice, pending funding | Subsquid GraphQL, positions sorted by size |
| Jupiter | GET /v1/positions?walletAddress= | Side, size, collateral, entry, liquidationPrice, fees | /v1/top-traders, 50 by profit and 50 by volume, per market and week |
| Variational | None documented | The API is market statistics only | None found |
| GRVT | POST /full/v1/positions | 401 “You need to authenticate” | None found. Public trades carry no account |
| edgeX | Private account API | Documented as key-only | None found |
| Antarctic | Tried paths on its open API | “sign-error” on all but open interest | None found |
| QFEX | GET /positions | 401 “no authentication credentials” | None found |
Hyperliquid: everything, including the file of every ranked account
Hyperliquid’s documentation says its order books are fully on-chain, and that every order, cancel, trade and liquidation happens transparently. Our test agrees. The public leaderboard file returned 48,471 accounts, of which 2,768 showed $1M or more in account value, and any of them could be read directly. The same call with a dex setting read a position on the separate xyz market, so builder-run markets under HIP-3 are visible too.
One detail is easy to miss. In our own November 5 data, 1,536 of 5,060 whale positions (30.4%) had no liquidation price at all, and 1,531 of those were cross margin. So “Hyperliquid shows liquidation levels” is true, but about three in ten large positions we track have an empty field at that moment. A chart of whale liquidation levels leaves those positions out.
Pacifica: complete, with one quirk in liquidation prices
Pacifica returned a ranking of 6,894 accounts with profit by period, equity and open interest, and positions for any address. Liquidation prices need care: one long we read carried a negative liquidation price, a level the market cannot reach. Treat that as “no practical level”, not as a number to average.
GMX: more than many guides say, split across networks
GMX positions are readable by address from its GMX API, and the response includes a computed liquidationPrice. GMX values use 30 decimals, so divide by 1030. Rankings are not a published file in what we tested; we built one by sorting account statistics by realized profit in its Subsquid GraphQL endpoint, so it scores 1.
The test applies to Arbitrum. GMX’s documentation covers only the pair feed on its Solana host, so we did not test wallet reads there. Since Solana holds most of GMX’s activity (see below), this is a large caveat: we did not confirm wallet-level reads for most of GMX’s open interest.
Jupiter: wallet reads are simple, market totals are not
Jupiter’s perps API returned positions, liquidation price and a wallet’s trade history. It also has a weekly top-traders call, limited to 50 wallets per list. Its SOL, ETH and wBTC markets trade against the JLP pool and charge a borrow fee that accrues continuously rather than funding, so there is no funding rate to read on those markets. We found no direct open interest figure in the calls we tried: DefiLlama’s adapter derives it from an on-chain query through Dune. Beta markets on a separate order book engine use hourly funding, and we did not test those.
Variational, GRVT, edgeX, Antarctic and QFEX
Variational says plainly that its API is public, read-only market statistics. It does give something others do not: long and short open interest for each listed market, in a venue that says it has 500+ live markets. We found no route to another user’s positions. GRVT gives a market ticker with open interest and a long/short ratio, and its positions call refuses unauthenticated requests. Its documentation site returned a 403 and its website showed a restricted-IP notice, so its score is the least certain among these. edgeX documents account data under a key-only private API; a funding request with our contract ID came back empty. Antarctic returned open interest and rejected our other requests with a signature error. QFEX returned open interest in the feed it provides to DefiLlama, and its positions path refused the request.
What do Lighter, Aster and dYdX publish, by their own documents?
We did not call these three markets, so nothing below is a test result and none of them has a score. It is what their public documentation and terms state, read on November 5, 2026. A documented field or call may behave differently in practice.
From each market’s documentation. Not tested, not scored.
| Market | A wallet’s positions | Finding large accounts | Liquidation price |
|---|---|---|---|
| Lighter | Account call labeled a public snapshot, with collateral, positions and sub-accounts | Public trade channel shows ask and bid account IDs. No leaderboard endpoint in the pages we read | Not confirmed in the pages we read |
| Aster | Position call documented as signed | None documented. Public trades response has no address field | None documented for other accounts. Liquidation orders appear on public streams, as events |
| dYdX | Position calls by address and subaccount number, with no authentication stated on that page | No ranking call in the indexer client page | No such field in the documented position object |
Sources: Lighter account data, Lighter WebSocket, Aster API, dYdX indexer, dYdX position object.
Lighter: documented as a public account view
Lighter’s integration guide labels the account call a public snapshot of collateral, positions and sub-accounts. Active orders, profit and history endpoints are documented as needing an auth token. Its WebSocket documentation shows ask and bid account IDs on each public trade. On paper, that is a route to active accounts without a ranking.
dYdX: position calls documented by address
dYdX’s architecture page describes a read-only indexer that serves chain data to users over REST and WebSocket, and its client documentation lists position calls by wallet address and subaccount number. The documented position object includes side, size, entry price and realized and unrealized profit, with no liquidation price field. Its liquidation page shows how that level is calculated from equity and margin, so a reader would have to model it.
Aster: documented as private, with a way to hide positions
Aster’s API documentation marks position calls as signed, meaning they return only your own account, and its public trades response has no address field. Liquidation orders appear on public WebSocket streams, which show events rather than levels.
Aster also documents Hidden Order and Hidden Position: positions opened through hidden orders are, in its words, “excluded from all public data and portfolio analytics”. Whatever a third party shows as Aster whale data, this feature means a share of positions could be invisible by design, and an outsider cannot measure that share.
This connects to an earlier dispute. In October 2025 The Block reported that DefiLlama delisted Aster’s perp volume, after its head said Aster’s volume had begun to mirror Binance’s and that Aster does not make it possible to see who is making and filling orders. DefiLlama lists Aster again today. We did not find a statement on when or on what basis it returned, and the absence of order-level data is not evidence of wash trading, only a limit on checking it.
How big is each market, and why do sources disagree?
DefiLlama reported figures, data clock November 5, 2026 11:45 UTC
| Market | Volume 24h | Volume 30d | Open interest | Share of 30d volume | Volume 24h / OI |
|---|---|---|---|---|---|
| Hyperliquid | $15.57B | $236.58B | $7.76B | 35.8% | 2.0x |
| Lighter | $3.94B | $75.13B | $744.6M | 11.4% | 5.3x |
| Variational | $2.68B | $73.87B | $920.9M | 11.2% | 2.9x |
| Aster | $2.21B | $47.62B | $1.44B | 7.2% | 1.5x |
| GMX | $2.14B | $25.22B | $237.8M | 3.8% | 9.0x |
| edgeX | $792.7M | $18.70B | $175.9M | 2.8% | 4.5x |
| GRVT | $563.2M | $15.07B | $429.4M | 2.3% | 1.3x |
| Pacifica | $438.6M | $10.94B | $57.1M | 1.7% | 7.7x |
| QFEX | $559.2M | $10.17B | $175.0M | 1.5% | 3.2x |
| Antarctic | $587.3M | $9.04B | $360.0M | 1.4% | 1.6x |
| Jupiter | $319.8M | $5.31B | $41.7M | 0.8% | 7.7x |
| dYdX | $43.7M | $1.12B | $52.3M | 0.2% | 0.8x |
Page totals: $35.13B 24h volume, $660.67B 30-day volume, $13.76B open interest. Share is the market’s reported 30-day volume over the page total. Method: figures read from the page’s own data, for each parent protocol. Reported volume is not adjusted for wash trading or incentives.
Hyperliquid’s market share has no single answer
On one page, on one day, Hyperliquid is 56.4% of open interest, 44.3% of 24-hour volume and 35.8% of 30-day volume. Published figures add further spread. A piece dated October 7, 2026 worked out about 33% of 30-day perp DEX volume from DefiLlama data. CryptoSlate in October 2025 showed 25.5%, a year earlier and with no source named for the chart. Any of these can be correct for its window and its list of markets. A single “market share” number without those two details is not meaningful.
The same open interest can be counted twice
The largest original surprise in this test is a convention, not a venue. Open interest can count one side of each contract or both, and APIs do not agree. For the markets we tested, we compared the raw API total against DefiLlama’s figure:
Raw API total vs DefiLlama, November 5, 2026 (API read about 12:46 UTC)
| Market | Raw from the market’s API | DefiLlama | Note |
|---|---|---|---|
| Hyperliquid main markets | $10.83B | $7.76B | DefiLlama’s code halves Hyperliquid’s figure, calling it two-sided |
| Hyperliquid HIP-3 markets | $4.63B | Added in, also halved: ($10.83B + $4.63B) / 2 = $7.73B | |
| Variational | $1.81B headline; $906.5M market sum | $920.9M | Headline is twice the per-market long plus short sum; its docs call the per-market figures single-sided |
| edgeX | $325.8M | $175.9M | Adapter halves it |
| QFEX | $355.2M | $175.0M | Feed value for November 4; adapter notes both sides |
| Pacifica | $113.1M | $57.1M | Ratio about 2.0 |
| GMX, all networks | $237.1M | $237.8M | Long plus short; each position is against the pool, so counted once |
So DefiLlama’s published adapter code states that Hyperliquid’s API counts both sides, measured at “exactly 2x” a one-sided figure from CoinMarketCap, and halves it. We did not verify that claim independently. The practical point stands either way: a comparison that takes one market’s raw number and another’s one-sided number can be off by a factor of two, which is larger than most of the gaps it is trying to show. DefiLlama also lists tradeXYZ separately at about $4.46B, which matches the raw xyz figure of $4.45B, while that same open interest is inside Hyperliquid’s total at half weight. tradeXYZ is outside the $13.76B page total, so this is not a double count in that total, but the two rows use different conventions.
Volume has the same trap. Pacifica’s own 24-hour market sum was $872.1M against DefiLlama’s $438.6M, also about 2x. And DefiLlama now shows both reported and normalized 24-hour volume for several markets: Lighter is $3.94B reported and $2.95B normalized, Hyperliquid and Aster show the same number in both. We could not find a definition of “normalized” in its data glossary, so we report the numbers without interpreting them.
Volume and open interest tell different stories
The last column of the size table divides 24-hour volume by open interest. It runs from 0.8x at dYdX to 9.0x at GMX. High turnover does not prove wash trading, and low turnover does not prove its absence. It does tell you when volume and resting positions describe different markets. Fee and reward design plausibly feed into it. Lighter’s standard accounts pay no maker or taker fees on regular orders, and Variational’s Omni charges none. Pacifica distributes 10,000,000 points weekly, while saying self-trading earns none, and Variational also runs a points program. We have not measured how much volume these drive at any market, and we would not claim to.
GMX is mostly Solana now, and that surprised us
GMX’s own stats endpoint, read at 12:52 UTC, gave $1.98B of 24-hour perp volume and $237.1M of open interest across networks. Solana was $1.90B of that volume (95.8%) and $182.9M of open interest (77.1%). Arbitrum was $78.4M and $51.7M. Turnover was about 10.4x on Solana against 1.5x on Arbitrum. We did not determine why, and we make no claim about the quality of that volume. Anyone quoting GMX as an Arbitrum market is now describing the smaller part of it, including, to be plain, our own GMX pages.
How do the markets differ by design?
Design explains much of the transparency result. A market that runs its whole matching state on a public chain has less to hide by default. One that keeps its book off-chain must choose to publish.
Design, from each market’s documentation. Chain labels per DefiLlama where noted.
| Market | Trading model | Where it runs | Margin | When an account fails |
|---|---|---|---|---|
| Hyperliquid | On-chain order book | Own L1, HyperBFT consensus | Cross or isolated; some strict isolated | Market orders to the book, then a backstop liquidator vault at 2/3 of maintenance margin |
| Lighter | Order book | Ethereum Layer 2; DefiLlama lists zkLighter and Robinhood Chain | Shared account margin, or isolated per position | Pre-liquidation limits, partial liquidation at a computed zero price, then close-out with its LLP fund |
| Aster | Order book in Pro mode; other modes exist | DefiLlama lists it as off-chain | Cross or isolated in Pro mode | Forced liquidation when margin falls under maintenance, using mark price |
| GMX | Oracle-priced against pools | Arbitrum, Avalanche, MegaETH, Solana | Collateral per position | Below a 0.25%-1% collateral threshold, plus auto-deleveraging of profitable positions to protect the pool |
| dYdX | Order book | Own L1 on CometBFT and Cosmos SDK | Cross, plus isolated markets | Protocol-generated liquidation orders, with the insurance fund taking the resulting profit or loss |
| Jupiter | Oracle-priced against JLP; separate order book for Beta | Solana | Collateral per position on JLP markets | Oracle price hits a liquidation price that drifts as fees accrue |
| Pacifica | Order book | Uses Solana wallets; DefiLlama lists it as off-chain | Cross or isolated | Market liquidation in chunks, backstop at 2/3 of maintenance, then auto-deleveraging |
| Variational | Request-for-quote, with its OLP as sole counterparty | Arbitrum per DefiLlama | Single cross-margined account | User and OLP are both margined and can both be liquidated |
Sources: Hyperliquid, Lighter, Aster, GMX, dYdX, Jupiter, Pacifica, Variational. Summaries are simplified; read the source for parameters.
Two consequences matter for anyone reading whale data. First, on an oracle-and-pool market the other side of a large position is the pool, so a big long there says something different from a big long matched against another trader on a book. Second, a liquidation price is only a snapshot. GMX warns that its level is not static because fees move it, and Jupiter describes the same drift. On a cross-margin account the level also shifts with every other position the account holds. A field you read at noon may be wrong by evening without the trader doing anything.
Which markets bar US persons from their interfaces or APIs?
We read the terms each market publishes. This is a factual summary and not legal advice. We do not interpret how any clause applies to reading public data. Note what each document covers, since “interface” and “API” are different things.
US restrictions in published terms, read November 5, 2026
| Market | What the terms say on the US | What they cover | Dated |
|---|---|---|---|
| Hyperliquid | Barred: residents, located or incorporated in the US, plus Ontario | The website interface. Our read found no mention of the API in this clause | June 15, 2026 |
| Lighter | Barred, with others including Canada and the UK | Explicitly includes website, interface, protocol and API | December 29, 2025 |
| Aster | Barred, with others including Canada and the UK | Explicitly includes website, apps and APIs | February 25, 2026 |
| GMX | User confirms they are not a US Person | The GMX.io site and its app interface | August 20, 2026 |
| dYdX | Trading software ineligible for US use; VPN circumvention prohibited | dYdX software, incl. the chain and documentation | March 12, 2026 |
| Pacifica | Trading access prohibited; restricted by IP | Trading functionality | Undated docs notice |
| Variational | United States and US persons listed as restricted | Its products and services | Page updated April 20, 2026 |
| edgeX | Interface unavailable in the US and Canada | The interface | Not stated on the copy we read |
| Jupiter | Mobile app terms name the US as a prohibited locality | The mobile app. We could not retrieve current web or perps terms | September 21, 2024 |
| GRVT | Not read; its site showed a restricted-IP notice to our connection | Unknown | Unknown |
| Antarctic, QFEX | Could not locate terms | Unknown | Unknown |
Sources: Lighter, Aster, GMX, dYdX, Pacifica, Variational, edgeX. Hyperliquid’s and Jupiter’s terms are published on their own app sites, which we do not link. Terms change, so check the current version.
Enforcement showed up in the test too. During it, Hyperliquid’s site displayed a restricted-jurisdiction notice and GRVT’s site a restricted-IP notice. Neither was avoided. Market data from both continued to answer, which shows only how each is configured, and not what its terms permit.
What can’t this test tell you?
- It is one connection on one morning. Geo-rules, endpoints and terms change, and a gap above can be the direct result of being blocked.
- “Not found” is weak evidence. A market may publish a ranking on a web page we could not load, or through a call we did not try. We would welcome corrections from anyone at these teams.
- Readable does not mean complete. A trader can split size across wallets or sub-accounts on any of these markets, so one address is not one person. We name no wallet owners, and an address on its own says nothing about who controls it.
- Public positions do not prove intent. A short may hedge a spot holding elsewhere. Nothing here tells you why a position exists.
- A score of 10 says nothing about risk. It does not measure solvency, security, oracle quality, fees or market fairness, and our ranking must not be read as a view on any of those.
What does Deep Blue Alpha cover, and what doesn’t it?
Our derivatives pages are built on the two markets where wallet reads worked and we had built collectors before this test: Hyperliquid from November 4, 2026 and GMX from November 5. History is days old, so we show snapshots and make no claims about trends or win rates. At 12:53 UTC on November 5 our data held:
Deep Blue Alpha snapshot, November 5, 2026 12:53 UTC
| Market | Rule for a whale | Accounts with open positions | Positions | Position value |
|---|---|---|---|---|
| Hyperliquid | $1M+ balance, has traded, excludes market makers | 662 | 5,060 | $5.91B |
| GMX (Arbitrum) | $100K+ in open positions; positions of $10K+ read | 98 | 235 | $38.7M |
There are real gaps, and a competing reader should know them. We cover Hyperliquid’s main markets only, which leaves out the $4.63B (raw) on HIP-3 markets such as xyz. We cover GMX on Arbitrum only: our $38.7M compares with $237.1M across GMX, or about 16%, because most GMX open interest sits on Solana. Our GMX pages do not yet show liquidation levels, even though this test found the GMX API returns them. We do not cover Pacifica or Jupiter, despite their high scores here, or Lighter, Aster and dYdX. Our older guide to whale derivatives covers Deribit and historic cases, and our Hyperliquid bridge analysis covers capital flows; neither repeats this test.
Bottom line
Asking which perp DEX is biggest and asking which shows you the most are different questions with different answers. On measured public data, Hyperliquid and Pacifica showed everything we looked for, GMX nearly everything, and Jupiter wallet-level data with thinner market data. Variational and GRVT showed market totals and no individual positions. We did not test Lighter, Aster or dYdX. Their documents describe public position reads at Lighter and dYdX, signed position calls at Aster, and a way for Aster positions to stay out of public data altogether.
Be careful with the numbers in between. A factor of two can come from nothing more than how open interest is counted, Hyperliquid’s share runs from 36% to 56% on one page, and GMX’s volume is now overwhelmingly on Solana. Before relying on any figure, ours included, ask which side was counted, which window and which list of markets. This ranking measures data availability only, and it is not a view on where anyone should trade.
See the positions this test could read
Biggest open longs and shorts, leverage, profit and latest moves on Hyperliquid and GMX, read about every 5 minutes from public data.
Open whale derivatives →