Decentralized perpetual exchanges are often grouped together as if they were interchangeable. They are not. Hyperliquid, dYdX, and GMX all let traders take leveraged long or short exposure without using a traditional centralized exchange account, but they differ sharply in how orders are matched, how liquidity is supplied, how fees accumulate, and what kinds of risks matter most.
To make those differences concrete, this article uses one clearly hypothetical example throughout. Imagine Maya, a trader with 5,000 USDC who wants to open a 10,000 USDC ETH perpetual position and may keep it open for roughly 24 hours. She is not trying to maximize leverage. Her priorities are execution quality, predictable costs, transparent risk, and the ability to exit quickly if the market moves against her. This is an illustrative scenario only, not a real trade, backtest, recommendation, or claimed result.
As of September 16, 2026, the most important distinction is architectural. Hyperliquid uses fully onchain order books inside HyperCore, its specialized trading execution layer. dYdX also uses an order-book model on its own blockchain, with validators participating in order handling and price-oracle processes. GMX does not use a traditional order book for perpetual execution; it routes trades against liquidity pools and uses oracle-based pricing with explicit mechanisms for price impact, funding, and borrowing.
| Category | Hyperliquid | dYdX | GMX |
|---|---|---|---|
| Core execution model | Onchain central limit order book | Order book on dYdX Chain | Oracle pricing against liquidity pools |
| Best fit in our example | Trader prioritizing fast order-book execution | Trader who wants order-book trading with chain-level governance | Trader who prefers pool-based execution and can monitor borrowing/funding costs |
| Fee structure | Maker/taker tiers based on volume, with staking discounts and possible rebates | Maker/taker schedule governed by protocol parameters | Position fee plus funding, borrowing, price-impact effects, and network execution costs |
| Collateral / margin | USDC-based perpetual margining; cross and isolated modes | Market-specific margin parameters; chain software supports isolated markets as well | Flexible collateral depending on market and network |
| Liquidity source | Resting orders from traders and market makers | Resting orders from traders and market makers | GM/GLV liquidity pools |
Action for Maya: before comparing headline fees, she should first decide whether she wants an order book or an oracle-and-pool execution model. That single choice changes how she should think about slippage, depth, price impact, and ongoing position costs.
Hyperliquid describes itself as a layer-one blockchain optimized for trading. Its HyperCore execution state contains fully onchain perpetual and spot order books, and orders are matched by price-time priority. The official documentation says every order, cancellation, trade, and liquidation is processed transparently through the chain’s trading system. See the Hyperliquid documentation and its order-book documentation.
For Maya, that means the execution mental model will feel familiar if she has used a centralized futures exchange: she can look at bids and asks, submit market or limit orders, and care directly about spread and displayed depth. This is different from GMX, where a trade is not consuming a conventional bid/ask book.
Hyperliquid’s official fee page currently shows a base perpetual fee tier of 0.045% for takers and 0.015% for makers before volume-based reductions, staking discounts, or maker rebates. Fees depend on rolling 14-day weighted volume, and the documentation also lists HYPE staking discounts. Because fee schedules can change, traders should verify the live tier rather than treating these numbers as permanent. The current schedule is published in the Hyperliquid fee documentation.
If Maya crossed the spread to open a 10,000 USDC position at the base 0.045% taker rate, the opening trading fee alone would be about 4.50 USDC. Closing as a taker at the same base rate would add another roughly 4.50 USDC. This simple calculation excludes funding, execution-price movement, and any future fee changes. It is an illustration of fee arithmetic, not a forecast of her final cost.
Common misunderstanding: “onchain” does not mean there is no exchange-style microstructure. Hyperliquid still has order-book depth, maker/taker behavior, funding, margin rules, liquidations, and market-specific constraints. Action: Maya should inspect both fee tier and visible depth for ETH before placing a large market order.
Hyperliquid’s contract specifications describe linear perpetuals with USDC margining and hourly funding payments. Cross margin and isolated margin are supported. Market leverage limits and order-size rules vary by asset. The relevant source is the Hyperliquid contract-specification documentation.
For a 24-hour hypothetical position, Maya should therefore compare more than entry and exit commissions. A seemingly cheap trade can become more expensive if funding is persistently unfavorable. Action: she should record the current funding rate when entering and set a threshold at which the carrying cost makes the trade unattractive.
dYdX also uses an order book, but its implementation and governance model differ from Hyperliquid. The dYdX Chain software supports market, limit, stop market, stop limit, take-profit market, and take-profit limit orders. The official help center distinguishes short-term orders, which can remain in validators’ memory for a limited block window unless filled, from longer-lived stateful orders that are committed to chain state. See the official dYdX order-type documentation.
For Maya, the practical consequence is familiar order-book trading with protocol-specific behavior under the hood. She can still think in terms of makers, takers, limit prices, time-in-force settings, and available market depth.
The dYdX help center explains that the default open-source software uses a maker-taker fee model, with taker tiers based on trailing 30-day USD trading volume. It also emphasizes that governance can adjust these parameters. The documentation states that canceled orders are not charged trading fees and that traders do not, by default, pay a separate gas fee for each trade in the same way they might on a general-purpose chain. Current details are available in the official dYdX trading-fee article.
Common misunderstanding: a decentralized exchange does not necessarily make users pay a visible blockchain gas charge on every order. dYdX’s software can abstract that experience into trading fees. Action: Maya should check her live fee tier in the interface rather than comparing an old fee table copied from another website.
dYdX’s funding mechanism uses a premium derived from impact bid and ask prices relative to the index price. The software calculates samples over time and applies funding on an hourly schedule. The official explanation is in the dYdX funding documentation.
In Maya’s 24-hour example, that matters because the exchange with the lowest entry fee may not be the cheapest venue over an entire day. Action: she should compare expected funding over her intended holding period, not just the fee charged when the position opens.
dYdX markets have initial and maintenance margin fractions, open-interest caps, and market-specific maximum leverage. The software can increase effective margin requirements as open interest rises beyond configured thresholds. That makes the maximum advertised leverage less informative than the live effective requirement in a stressed market.
Action: Maya should check the selected ETH market’s current margin parameters and leave a meaningful equity buffer instead of sizing her position right up to the liquidation boundary.
GMX requires a different mental model. Its current documentation describes a decentralized spot and perpetual exchange operating on supported networks such as Arbitrum and Avalanche, with trading powered by GM and GLV liquidity pools. Orders are priced using oracle data rather than matched against a conventional central limit order book. The protocol documentation is available at the official GMX documentation.
For Maya, this means she should not ask, “How deep is the ETH bid book at my price?” in the same way she would on Hyperliquid or dYdX. Instead, she should look at pool capacity, open-interest imbalance, oracle pricing, price-impact rules, borrowing rates, funding, and execution fees.
The GMX fee documentation currently states that most markets charge a position fee of 0.04% or 0.06% when increasing or decreasing a position, depending on whether the trade improves or worsens the imbalance between long and short open interest. GMX also has funding fees, borrowing fees, network execution costs, and price-impact mechanics. See the current GMX fee documentation.
If Maya opens a 10,000 USDC position, a 0.04% position fee would equal 4 USDC and a 0.06% fee would equal 6 USDC for that size change. But treating that as the “total GMX fee” would be misleading because borrowing and funding can accumulate while the position remains open, and price-impact effects can matter when the position is reduced or closed.
Common misunderstanding: GMX is not simply “an AMM with a fixed swap-style fee.” Its perpetual system has several dynamic cost components tied to open-interest balance and pool utilization. Action: Maya should inspect the live borrow rate, funding rate, expected position fee, network fee, and estimated price impact together before entering.
GMX uses Chainlink Data Streams and protocol-specific execution logic for pricing. Its order documentation explains the use of minPrice and maxPrice and how different actions reference those bounds. Oracle pricing can reduce dependence on a locally visible order book, but traders still face price movement, pool constraints, keeper/execution timing, funding, borrowing, and smart-contract risk. See the GMX positions and order-types documentation.
Action: Maya should read the estimated execution values shown before signing and avoid assuming that “no order book” means “no slippage or price impact.”
Suppose all three venues list the ETH perpetual she wants and Maya is comfortable with the relevant chain and wallet setup. She can now reduce the decision to four practical questions.
If yes, Hyperliquid and dYdX are the natural first comparison. She can inspect displayed liquidity, use limit orders, and reason directly about maker versus taker execution. Hyperliquid may appeal if she prioritizes a trading-first L1 with an onchain order book and highly integrated trading infrastructure. dYdX may appeal if she values its own chain architecture, governance-adjustable market parameters, and familiar advanced order types.
Action: compare the actual ETH spread, depth, funding, and personal fee tier on both venues at the moment of the trade.
If yes, GMX is structurally different enough to deserve separate analysis. A trader who dislikes worrying about every individual order-book level may prefer an oracle-and-pool framework, but that does not remove cost uncertainty; it shifts the important variables toward pool balance, borrowing, funding, and price-impact rules.
Action: estimate the full round-trip cost under the current long/short imbalance rather than comparing only the 0.04% or 0.06% position fee.
For very short trades, spread and entry/exit fees can dominate. For positions held many hours or days, funding and borrowing become more important. Hyperliquid and dYdX both use funding to keep perpetual prices near underlying references. GMX adds borrowing costs tied to pool utilization in addition to funding mechanics.
Action: write down an expected holding period before opening the position, then compare carrying costs over that same horizon.
During a sharp move, the exchange’s architecture matters. Order-book venues can experience rapidly changing spreads and depth. Pool-based venues can experience changing imbalance, borrowing rates, and price-impact conditions. All three can liquidate undercollateralized positions according to their respective rules.
Action: Maya should size the trade based on the loss she can tolerate, not the maximum leverage the interface allows. A 2x position in this example leaves substantially more room for adverse movement than using the same 5,000 USDC to chase the platform’s maximum leverage.
There is no defensible universal winner because the platforms optimize different things.
Those are use-case recommendations, not guarantees about returns, liquidity, or future protocol quality. All three systems can change through software upgrades, governance decisions, market listings, fee revisions, and risk-parameter updates.
Verified from current official documentation: Hyperliquid and dYdX use order-book-based perpetual trading; GMX uses oracle pricing against liquidity pools. Hyperliquid publishes maker/taker fee tiers, dYdX uses governance-adjustable maker/taker parameters, and GMX has position fees plus funding, borrowing, and execution-related costs.
Condition-dependent: the cheapest venue for Maya’s example depends on her personal fee tier, whether she is maker or taker, market depth, funding direction, pool imbalance, borrowing utilization, holding period, network conditions, and the size of her order.
Not knowable in advance: which exchange will have the best effective execution at the exact moment she trades, how funding will evolve over the next 24 hours, or which platform will experience future outages, governance changes, software bugs, oracle incidents, liquidity shocks, or smart-contract failures.
Action: before every material trade, reopen the official fee and risk pages, verify the live market conditions in the trading interface, and compare the estimated total cost of the complete position lifecycle—entry, holding period, and exit.
For Maya’s hypothetical 10,000 USDC ETH perpetual, the right choice is therefore not the exchange with the most impressive headline or the lowest isolated fee number. It is the venue whose execution model, live liquidity, carrying costs, and risk controls best fit how she actually intends to trade.
Learn how to evaluate Web3 gaming tokens by emissions, token sinks, unlocks, retention quality, and sustainable player demand—not just headline rewards.
Learn how wallet drainers exploit phishing, token approvals, NFT operator permissions, and signed permits—and how to verify requests, limit access, revoke risk, and respond to compromise.
Compare six AI-focused Layer 1 blockchains by what they actually optimize for: agents, compute, data, on-chain inference, interoperability, and developer fit.
Compare Banana Gun, Maestro, and Sol Trading Bot through a hypothetical SOL trade, with practical wallet, Telegram, slippage, fee, and scam-safety checks.
A practical guide to tokenized private equity and private debt: how it works, real-world examples, benefits, risks, regulation, and what to watch next.
Compare Hyperliquid, dYdX, and GMX by execution model, fees, funding, liquidity, collateral, and risk using a practical hypothetical trading scenario.
Understand modular vs. monolithic blockchains, where Celestia and EigenLayer fit, the trade-offs in security and scalability, and what may shape crypto next.
Practical crypto derivatives risk rules for leverage, position sizing, stops, liquidation, margin, funding, correlation, and exchange risk.
A beginner-friendly guide to restaking with EigenLayer and Symbiotic: how rewards are created, where slashing risk comes from, and how to compare opportunities.
Learn how Bitcoin Ordinals and BRC-20 tokens work, why they create demand for blockspace, how they can affect fees, and what beginners should check before using them.
