Ranked & Reviewed · 9 services
Best Perpetual DEXs
Nine on-chain perpetual futures venues, scored on published fees, funding and liquidation terms, core code licence, audits and risk documentation, from primary sources checked for November 2026.
At a glance
| # | Service | Score | Best for |
|---|---|---|---|
| 1 |  dYdX | 8.6 | Open AGPL core with fee and liquidation rules in numbers |
| 2 |  Ostium | 8.6 | Audited, MIT-licensed contracts with no criterion below 8 |
| 3 |  GMX | 8.4 | Documented fees and funding, with a long named audit record |
| 4 |  Orderly Network | 7.4 | Funding and margin rules in full, backed by 17 public audit reports |
| 5 |  Paradex | 7.2 | Fee, funding and liquidation rules published as figures |
| 6 |  Hyperliquid | 7.0 | Funding and liquidation terms documented to the exact formula |
| 7 |  Aevo | 6.0 | Fees, funding and liquidation terms published as figures |
| 8 |  Gains Network | 5.8 | Per-asset fees and liquidation thresholds published in full |
| 9 |  Jupiter Perps | 5.8 | Itemized trading costs and a named audit record for the program |
Best Perpetual DEXs, ranked by score
dYdX scores 8.6 among perpetual DEXs, tied with Ostium and listed second under the alphabetical tie rule. It has an AGPL-licensed core, a fully tabled fee schedule and a 1.5% default liquidation penalty paid to an insurance fund. Its main limits are an audit record from a single firm ending in 2024 and an undisclosed insurance fund size, according to desk research checked November 1, 2026.
Pros
- Core v4-chain code licensed under AGPL-3.0, covering the matching and liquidation logic (licence mark 10)
- Seven fee tiers published, from 5.0 bps taker and 1.0 bps maker at tier 1, plus the staking discount table
- Hourly funding formula and 8-hour cap rule documented, with margins by market tier
- Six Informal Systems audit reports, 2023 to 2024, linked from the docs Security page (audit mark 7)
- Default 1.5% maximum liquidation penalty, paid to an insurance fund, with deleveraging explained by example
Cons
- All six published audit reports come from one firm, Informal Systems, and none is newer than 2024 (audit mark 7)
- Insurance fund size is not published in the documentation
- Fee, funding and liquidation values are defaults that governance can change
- Licence carries an extra condition that ends rights for anyone breaking applicable law or the terms
Ostium scores 8.6, the top mark among perpetual DEXs, with six dated audit reports, MIT-licensed core contracts and a single published liquidation formula. Its rollover fee is harder to reproduce than standard funding. Marks come from desk research of its docs, GitHub and contracts, checked November 1, 2026.
Pros
- Six audit reports with scope and findings, February 2024 to January 2026 (audit mark 10)
- Core contracts MIT-licensed and updated to v1.5.0 in May 2026 (licence mark 9)
- Opening fee of 3 to 10 bps by pair, with no closing or separate liquidation fee
- Liquidation follows one published formula with a 25% collateral backstop
- Upgrades pass through a timelock with an 18-hour minimum delay, verified on-chain
Cons
- All remaining collateral is retained on liquidation, with nothing returned
- Rollover carry method is published, but the 1% to 2% markup is not given per pair
- The junior buffer that shields the vault has no stated size
- SDK repositories carry no licence; oracle and hedging services are not public code
GMX scores 8.4 among perpetual DEXs, third of nine on current peer scores. It earns 9s for fees, funding and liquidation terms and audit disclosure, and its licence has now passed its change date to GPL v2.0 or later. The main limit is the lack of any insurance fund beyond its pools, according to desk research checked November 1, 2026.
Pros
- Position fee published as 0.04% or 0.06% depending on the imbalance a trade creates (fee mark 9)
- Adaptive funding and borrow-fee formulas documented, with typical utilization parameters
- Liquidation threshold and fees by market type published: 0.20%, 0.30% and 0.45%
- Code now granted under GPL v2.0 or later by the BUSL file's own terms, after its August 31, 2026 change date
- Five audit firms named, with Guardian engagements from 2022 to 2026 and an Immunefi bounty (audit mark 9)
Cons
- No insurance fund; liquidity providers and, through ADL, profitable traders absorb shortfalls (risk mark 7)
- LICENSE file still displays the BUSL text after its change date, and GitHub reports no recognized licence (licence mark 8)
- Live funding and borrow rates vary by market and appear in the interface, not in a table
- Price impact on close can reach 10% on thin markets
Orderly Network scores 7.4 among perpetual DEXs, with a full funding and margin specification and 17 public audit reports. Its matching and risk engines are closed and the fee a trader pays depends on the front end. Marks come from desk research of Orderly's docs and GitHub, checked November 1, 2026.
Pros
- Funding formula with a per-market interval, cap and floor table (funding mark 9)
- Margin formulas and per-market parameters; liquidation fees of 0.60% and 1.20% published
- 17 audit reports from eight sources, July 2022 to November 2025, in a public repository
- Liquidation flow, insurance fund and ADL triggers documented (risk mark 7)
Cons
- Matching and risk engines run off-chain and are not published (licence mark 5)
- The fee a trader pays is set by each front end and cannot be read from Orderly's docs (fee mark 7)
- Insurance fund size and its solvency parameter values are not disclosed
- Upgrade multisig signers and threshold are not named
Paradex scores 7.2 among perpetual DEXs, fifth of nine, on 9s for fee transparency and for funding and liquidation clarity. Its main limitation is closed core code: no public repository holds its clearing contracts. Re-scored from 2.8 on desk research checked November 1, 2026.
Pros
- Fee transparency 9: retail orders pay 0%; pro fees tabled at 0.003% maker, 0.035%-0.045% taker, 0.0175% floor
- Funding and liquidation 9: funding formula, caps and a 70%-of-maintenance-margin penalty stated as numbers
- Audit disclosure 8: dated reports from Cairo Security Clan, Zellic, Trail of Bits and HackerOne
- Risk documentation 8: insurance fund, socialized-loss factor and last-resort mechanism explained
Cons
- Licence 2: no public repository holds the core Paraclear contracts; only SDKs and a bridge fork are open
- Only one audit, from May 2025, covers the core contracts; the others review the bridge
- Insurance fund size is not published, and losses beyond it fall on users who withdraw
- Operator controls can block deposits and limit withdrawals
Hyperliquid scores 7.0 among perpetual DEXs, with the category's only 10 for funding and liquidation clarity and a fully tabled fee schedule. Its closed matching engine and an audit record limited to a legacy bridge hold it back. Marks come from desk research of its documentation and GitHub, checked November 1, 2026.
Pros
- Funding formula, clamp, 4% hourly cap and impact notional all published (funding mark 10)
- Exact liquidation-price formula and a 1.25% to 16.7% maintenance margin band
- Full perps fee table: 0.045% taker / 0.015% maker at base, seven tiers and staking discounts
- Liquidation sequence, backstop vault and ADL ranking index written out (risk mark 8)
Cons
- Matching engine is closed; open licences cover only SDKs and node setup files (licence mark 2)
- Bridge2.sol is marked UNLICENSED and the contracts repository has no licence file
- Zellic's two reports cover only the legacy Arbitrum bridge, not the exchange core (audit mark 6)
- Builder fees of up to 0.1% on third-party front ends sit outside the base fee table
Aevo scores 6.0 among perpetual DEXs because it publishes its fees, hourly funding formula, liquidation tiers and margin levels as numbers. Its closed off-chain core pulls the score down most, and its Veridise and Spearbit audit reports sit in GitHub without any mention in the docs. The marks come from desk research of Aevo's documentation and GitHub, checked November 1, 2026.
Pros
- Standard perpetuals fees published as 0.08% taker and 0.05% maker on notional (fee mark 8)
- Hourly funding with the full formula, clamp and per-asset caps documented
- Liquidation fees listed as 1%, 3% and 5% by tier, with margins stated per contract
- Liquidation sequence, insurance-fund step and ADL ranking formula all written out (risk mark 7)
Cons
- Off-chain matching and risk engine; only SDKs and tooling are public code (licence mark 2)
- Audit reports by Veridise (2023, marked draft) and Spearbit (2022) sit only in GitHub repositories; the docs name no auditor (audit mark 5)
- Insurance fund size and funding source are not disclosed
- Fee tiers beyond the standard rate are not listed, and a notice says fees will change
Gains Network scores 5.8 among perpetual DEXs. It publishes one of the most complete fee schedules in the category and a full liquidation-threshold table, but its core contracts are not public and its funding rate has no formula. Only one named audit, by Pashov Audit Group, can be reached from its docs, according to desk research checked November 1, 2026.
Pros
- Opening and closing fees listed per asset class, from 0.012% forex majors to 0.06% non-core crypto (fee mark 9)
- Liquidation thresholds tabled by leverage and asset class, with a liquidation-price formula and example
- Borrowing fee published as an explicit per-block formula
- Vault upgrades that could affect funds pass through a 14-day timelock
Cons
- Core gTrade-contracts repository is not public; only SDKs and tooling are open (licence mark 3)
- No audits page; the one reachable review (Pashov) is undated, and the Halborn link returns 404 (audit mark 5)
- Funding rate described without a formula or parameters
- Backstop relies on minting GNS, capped at 0.05% of supply a day, with no insurance fund or circuit breaker
Jupiter Perps scores 5.8 among perpetual DEXs: its fees are itemized with formulas and three named audits cover its program, but the program is closed source and liquidation forfeits all remaining collateral on the main markets. Marks come from desk research of Jupiter's documentation and GitHub, checked November 1, 2026.
Pros
- 0.06% base fee to open and close, with base plus price impact fees capped at 0.50% (fee mark 8)
- Price impact and borrow fee formulas published with the on-chain accounts that hold their parameters
- Liquidation-price formula published, with a warning that borrow fees move it over time (funding mark 7)
- Three named audits with linked reports: OtterSec, Sec3 and Offside Labs (audit mark 7)
Cons
- Perpetuals program source is not published by Jupiter (licence mark 1)
- On SOL, ETH and wBTC markets all remaining collateral is lost on liquidation
- No insurance fund or bad-debt rule described beyond the JLP pool
- Audits date from 2023 to February 2024; the program has been upgraded since
Best perpetual DEXs by use case
- You want to read the code that matches and liquidates your trades: dYdX is the only venue with a 10 for licence. Its v4-chain repository, which holds the matching and liquidation logic, is under AGPL-3.0, with an added condition that ends rights for anyone breaking the law or its terms. Its audit mark is 7 because all six reports come from one firm, Informal Systems, and none postdates 2024.
- You want the deepest audit record with no weak criterion: Ostium has the only 10 for audits, from six reports by Zellic, ThreeSigma and Pashov Audit Group dated 2024 to 2026, and no mark below 8. Two things to know first: it charges a rollover fee instead of long/short funding, and a liquidated position keeps none of its remaining collateral.
- You want to work out your own liquidation price and funding cost exactly: Hyperliquid holds the category's only 10 for funding and liquidation clarity. It publishes the funding formula with a 4% hourly cap, the exact liquidation-price formula and maintenance margins from 1.25% to 16.7%. Its matching engine is closed, so its licence mark is 2.
- You are a retail trader and fees are your main cost: Paradex has charged retail orders 0% maker and taker since 16 September 2025, and tables its pro tiers in full (fee mark 9). Its core clearing contracts are not in any public repository, and its restricted-persons page excludes US and Canadian persons.
- You prefer trading against a liquidity pool with a long audit list: GMX scores 9 for fees, for funding and liquidation terms and for audits, with five named firms. Position fees are 0.04% or 0.06% depending on whether a trade reduces or adds to the long/short imbalance. It has no insurance fund beyond its pools.
How the ranking works
Each venue is marked from 0 to 10 on five criteria, read from its own documentation, repositories and contracts:
- Fee transparency: maker and taker rates at the entry tier published as figures before you open a position.
- Funding & liquidation clarity: the funding mechanism and interval, and the liquidation penalty or maintenance margin, stated as numbers rather than left to the position screen.
- Licence & code openness: the licence on the core protocol, not the SDK. An open SDK wrapped around a closed matching engine scores low; a half-open core, such as Orderly's on-chain contracts beside an unpublished off-chain engine, sits mid-table.
- Audit disclosure: reports published with firms named and dated. An unreachable report counts as absent.
- Risk documentation: the insurance fund or backstop, the maintenance-margin schedule and how bad debt is handled.
The overall score is the plain average of the five marks, equally weighted and rounded to one decimal. Ties are listed alphabetically: dYdX sits above Ostium at 8.6, and Gains Network above Jupiter Perps at 5.8. All marks come from desk research of primary sources checked on 1 November 2026, with no trading or test positions. Marks are re-checked when documentation changes, and each review records what moved and why. The rules for every category are on the ratings methodology page.
| Venue | Fees | Funding & liquidation | Licence | Audits | Risk docs | Overall |
|---|---|---|---|---|---|---|
| dYdX | 9 | 9 | 10 | 7 | 8 | 8.6 |
| Ostium | 8 | 8 | 9 | 10 | 8 | 8.6 |
| GMX | 9 | 9 | 8 | 9 | 7 | 8.4 |
| Orderly Network | 7 | 9 | 5 | 9 | 7 | 7.4 |
| Paradex | 9 | 9 | 2 | 8 | 8 | 7.2 |
| Hyperliquid | 9 | 10 | 2 | 6 | 8 | 7.0 |
| Aevo | 8 | 8 | 2 | 5 | 7 | 6.0 |
| Gains Network | 9 | 7 | 3 | 5 | 5 | 5.8 |
| Jupiter Perps | 8 | 7 | 1 | 7 | 6 | 5.8 |
The licence column decides most of the order. Fee and funding marks sit between 7 and 10 for every venue, but five of the nine score 3 or less for code openness because their core engine or program is not public. Hyperliquid, Paradex and Aevo would rank higher on documentation alone.
What changed in this update
The November 2026 re-check raised most scores as venues published pages that were missing at the July check:
- Paradex rose the furthest, to 7.2 from 2.8 in July. Its funding, liquidation, audit and socialized-loss pages, previously unreachable, now state figures, and its fee page now tables the pro tiers.
- Jupiter Perps now has 5.8, up from 4.6, after its security page listed three audits of the perpetuals program with linked reports (audits 2 to 7) and its liquidation page published the price formulas.
- Orderly Network gained on funding and liquidation clarity (3 to 9) and risk documentation (3 to 7) as its funding, margin, liquidation and insurance-fund pages appeared, but lost on licence (9 to 5): its orderbook design page says matching and risk run off-chain in code that is not public.
- dYdX moved from 3 to 7 on audits once its Security page was found to link six Informal Systems reports. GMX moved from 5 to 8 on licence because its BUSL change date of 31 August 2026 has passed, converting the V2 code to GPL v2.0 or later by the file's own terms.
- Aevo and Gains Network each rose from 2 to 5 on audits, after Veridise and Spearbit reports were found in Aevo's GitHub and a Pashov Audit Group review was linked from the Gains docs.
- Hyperliquid slipped on licence, 3 to 2: the node repository credited as Apache-2.0 holds only setup files, and the node itself is a closed binary. Ostium was unchanged.
Drift has been withdrawn from the ranking: it was exploited on 1 April 2026 and relaunched as Velocity.
What the scores do not cover
- Leverage risk itself. A clear liquidation formula tells you when you will lose the position; it does not make losing less likely. On Ostium and on Jupiter Perps' SOL, ETH and wBTC markets, liquidation takes all remaining collateral.
- Depth and execution. Order-book depth, price impact and oracle behavior under stress are not scored. GMX's docs say price impact on close can reach 10% on thin markets.
- Backstop size. Several venues explain their insurance fund but not its size: dYdX, Paradex, Aevo and Orderly do not publish one. GMX and Gains have no insurance fund, and Gains' backstop relies on minting its own GNS token.
- Incident history. GMX V1 was exploited on 9 July 2025, when about $40 million of GLP liquidity was withdrawn. Audits and documentation marks do not price that in.
- Who may use the venue. Access rules are not scored and vary widely: Hyperliquid's interface terms exclude US and Ontario residents, Ostium's bar the US, the UK and the European Union, and Orderly, Paradex and Jupiter exclude US persons.
- Front-end fees. On Orderly and Hyperliquid, third-party front ends can add their own fee on top of the published base rate.
What is the difference between an order-book and a pool-based perpetual DEX?
On an order-book venue such as dYdX, Hyperliquid, Orderly or Aevo, your trade is matched against other traders' orders, and an insurance fund or auto-deleveraging closes out losses beyond a liquidated account. On a pool-based venue such as GMX, Jupiter Perps, Gains Network or Ostium, you trade against a liquidity pool or vault, so its depositors are your counterparty and, on GMX and Gains, absorb shortfalls in place of a separate fund. The holding cost differs too: Jupiter Perps charges an hourly borrow fee instead of funding, and Ostium a rollover fee.
How to choose a perpetual DEX
- Confirm access first. Read the interface's terms and restricted-persons list for where you live.
- Price the full cost of holding. Add the taker fee on entry and exit to funding, borrow or rollover charges for the time you expect to hold. Our guide to maker, taker and gas fees covers the trading side.
- Find the maintenance margin for your market and leverage, and what share of collateral a liquidation takes.
- Check what stands behind the venue: an insurance fund and its size, a pool, or auto-deleveraging that closes profitable traders' positions.
- Check the version audited. Jupiter Perps notes its program has been upgraded since its last audit in February 2024.
- Size positions so a liquidation is survivable. Our risk management guide explains position sizing and loss limits.
How we score this category
A perpetuals venue is judged on what happens to a position under stress, not on advertised leverage, so this rubric scores the published cost of trading, the funding mechanism, the liquidation terms and the code licence. Where a protocol's core matching engine is closed while only its SDK is open, we say so — an open SDK is not an open exchange.
Every service in this ranking is scored 0–10 against the same 5 criteria, each carrying equal weight (20% apiece). The overall score is their plain average, rounded to one decimal place, and each review prints the individual numbers and the arithmetic so you can reproduce the result yourself.
What each criterion covers
- Fee transparency20% of the score
- Whether maker and taker rates at the entry tier are published as figures a trader can look up before opening a position, rather than described only in general terms.
- Funding & liquidation clarity20% of the score
- Whether the funding mechanism, its interval and the liquidation penalty or maintenance margin are documented as numbers rather than left to be discovered on the position screen.
- Licence & code openness20% of the score
- The licence on the core protocol repository, read from the file — not the SDK. A recognised open-source core scores highest; a delayed-open licence mid-table; a closed core, or only an open SDK wrapping a closed engine, scores low.
- Audit disclosure20% of the score
- Whether audit reports are published with the firms named and, ideally, dated. An unnamed or unreachable audit record is scored as absent.
- Risk documentation20% of the score
- Whether the insurance or backstop mechanism, the maintenance-margin schedule and the treatment of bad debt are explained, so a trader can understand what stands behind the venue when a position blows through its collateral.
What does not affect a score
Only the criteria above count. No provider can buy a score, a rank, or an entry in this list. Where two services finish level, they are listed alphabetically rather than ordered silently.
The process across all categories, including how often we re-check, is set out in our ratings methodology.
Frequently asked questions
How were the perpetual DEXs in this ranking selected?
The ranking includes the nine on-chain perpetuals venues with a current Coin Currents Daily review, each marked on all five criteria from primary sources. Venues are not added to fill out a list. Drift was withdrawn after it was exploited on 1 April 2026 and relaunched as Velocity.
How is a perpetual DEX's overall score calculated?
Each of the five criteria is marked from 0 to 10, and the overall score is their equal-weight average, rounded to one decimal. dYdX, for example, has 9, 9, 10, 7 and 8, which averages 8.6. Ties are listed alphabetically.
Which risk does the score not capture?
Mainly the risk of leverage itself and the depth of the market under stress. A venue can document its liquidation rules perfectly and still liquidate you in a fast move. Insurance fund sizes are also often unpublished, so the score cannot say how much loss a venue can absorb.
Which perpetual DEX fits a trader on Solana?
Jupiter Perps is the Solana-native venue in this ranking, at 5.8. Its fees are itemized with formulas and three named audits cover its program, but the program source is not published and liquidation on its main markets forfeits all remaining collateral.
What fees and country restrictions should I check before trading?
Check the entry-tier taker and maker rates, any front-end fee, and the funding, borrow or rollover cost of holding a position. Then read the restricted-persons list: most venues here exclude US persons, and some also exclude the UK, the EU, Canada or Singapore.
How old are the measurements?
All marks were checked against primary sources on 1 November 2026; the previous check was in July 2026. Fee schedules and parameters on these venues can be changed by governance, so each review logs score changes with the source.
Rankings are editorial. Nothing here is financial advice. Editorial policy · How we score · How we're funded.