Global perp DEX competitive deep-dive matrix

Confidence: Likely Updated 2026-05-25 Review by 2026-11-25 Sources 12 Machine-translated Original (JA)
#exchanges#matrix#dex#perp#derivatives#defi
On this page

TL;DR

Perpetual-futures DEXs sit on three structurally distinct liquidity models — CLOB (central limit order book on-chain or hybrid), vAMM / pool-counterparty (single multi-asset LP pool serves all traders), and hybrid / oracle-pricing — and a reader who flattens them into “all perp DEXs” will misread fee, slippage, MEV resistance, LP risk, and regulatory exposure for every venue. This entry is the deep-dive 11-way matrix that complements the 5-way summary by adding the second generation of derivative DEXs (Synthetix Perps, Kwenta, Aark, Aevo, Apex) and the v1 / v2 split for protocols that materially rebuilt their architecture (dYdX v3 vs v4, GMX v1 vs v2). The cross-axes are chain, liquidity model, fee tier, max leverage, asset coverage, settlement asset, MEV protection, gas economics, daily-volume range, and perp-DEX market-share band. The matrix is read alongside the global spot DEX major-5 comparison and the native-chain-DEX flip-incumbent pattern to understand where on-chain derivatives sit relative to spot DEX share migration.

Wiki route

This sits under exchanges index as the perp-DEX deep-dive matrix. Read it with global perp DEX top 5 comparison (the predecessor that this entry expands), global DEX major five comparison for the spot DEX peer reading, native DEX flip-incumbent pattern for the chain-native vs cross-chain structural lens, Solana ecosystem DEX comparison for the Solana-side detail relevant to Drift, cross-chain bridge × CEX deposit/withdrawal for the settlement-asset on/off-ramp story, AMM design evolution for the LP-counterparty model history relevant to vAMM perps, liquid-staking / restaking CEX exposure for the collateral-side LST / LRT routes that several perp DEXs accept as margin, cross-chain five-pole comparison matrix for the broader chain landscape, and systems index for the cross-chain routing surface.

Why this matrix matters

By 2025-2026, perp DEXs collectively cleared meaningful single-day volumes that were occasionally comparable to mid-tier CEX peers, but the underlying architecture is so divergent that a fee-tier comparison alone is misleading. Hyperliquid’s purpose-built L1 CLOB has different incident-response properties from dYdX v4’s Cosmos-validator-coordinated CLOB; GMX’s GLP-pool counterparty model has different LP-risk economics from Drift’s vAMM + oracle pricing; Synthetix Perps’ Perps v2 / v3 design has different settlement timing from Kwenta’s front-end on Synthetix backend; Aevo, Apex, and Aark each chose different chain bases (Layer-2 vs OP Stack appchain vs StarkEx) for different reasons. The matrix below lays the architecture, gas, fee, leverage, asset coverage, MEV protection, and market-share axes side by side so a reader can pick the right venue for the right use case (high-leverage liquid pair vs long-tail asset vs MEV-sensitive flow vs LST-collateralised margin) rather than defaulting to whichever DEX has the highest 24h volume on a given day.

Per-DEX deep-dive

dYdX v3 (StarkEx L2, 2021-2023)

The original CLOB perp DEX at scale, running on a custom StarkEx instance (zk-rollup Layer-2 on Ethereum) with off-chain order book matching and on-chain settlement. Settlement asset: USDC. Max leverage: up to 20x on top pairs. Asset coverage: ~40 pairs at peak (majors + top-100 alts). MEV protection: high — off-chain matching engine with periodic L1 commits removes most of the on-chain MEV surface for individual trades. Gas economics: trader pays no gas (deposit/withdraw only), since matching is off-chain; deposit/withdrawal uses Ethereum-L1 gas via the StarkEx escape hatch. Fee tier: maker 0.02% / taker 0.05% standard. Daily-volume range (active period 2022-2023): commonly $0.5-2B. Market share: peak perp-DEX leader for the 2022 cycle.

The v3 architecture had limits: (a) reliance on the off-chain matching server centralises trust, (b) DYDX token governance was decoupled from protocol fees (no native trading-fee accrual to token holders), (c) the StarkEx system constrained chain-level extensibility. dYdX’s strategic answer was the v4 rebuild — see next row. Detail: cross-link to cross-chain bridge × CEX deposit/withdrawal for the USDC settlement-asset on/off-ramp.

dYdX v4 (Cosmos appchain, 2023-10 launch)

Full architectural rebuild on a custom Cosmos appchain (dYdX Chain) using CometBFT consensus. CLOB matching is now performed by validators (each validator runs its own order book and validators reach consensus on the canonical ordering), with all order-book state and trade settlement on-chain. Settlement asset: USDC. Max leverage: 20x. Asset coverage: 40+ pairs. MEV protection: high (validator-side matching with explicit anti-MEV mechanism). Gas economics: trader pays validator-network gas (paid via DYDX), but transaction cost is minimal. Fee tier: maker 0.02% / taker 0.05% with rebates. Daily volume: typically $0.3-1B (lower than the v3 era as Hyperliquid took share). Market share: top-3 perp DEX through 2024-2025.

v4 is the first fully-on-chain CLOB perp DEX at scale and the architectural template for the post-2023 generation. The trade-off: by moving matching on-chain into validator consensus, dYdX added the operational risk of validator-set capture but gained credible decentralisation. US-trader access is geo-blocked. Detail: cross-link to cross-chain five-pole comparison matrix for the Cosmos appchain context.

Hyperliquid (Hyperliquid L1, 2023-)

Purpose-built Layer-1 (HyperBFT consensus, custom EVM-compatible execution layer “HyperEVM”) with fully on-chain CLOB designed for low-latency order-book trading. Settlement asset: USDC (bridged from Arbitrum). Max leverage: up to 50x on majors. Asset coverage: 100+ perp pairs at any given time, with rapid listing of long-tail assets. MEV protection: high (custom L1 with order-flow design optimized against typical MEV vectors). Gas economics: trader pays HyperEVM gas, but cost is sub-cent per trade. Fee tier: maker rebates available, taker 0.025-0.05% tiered. Daily-volume range: $1-5B (occasionally higher during high-volatility events). Market share: dominant perp DEX leader by volume through 2024-2025, with HYPE token TGE 2024-11 distributed via no-pre-sale emission to retroactive users.

Hyperliquid’s distinguishing operational achievement is delivering CEX-class order-book UX (sub-second order placement, tight bid-ask spreads, deep books) on a fully on-chain venue. The team’s design choice — running everything on a custom L1 rather than using an existing L1 or L2 — is the key architectural bet. Detail: cross-link to native DEX flip-incumbent pattern for the chain-native flip analysis.

Vertex Protocol (Arbitrum, 2023-)

Hybrid CLOB + AMM design on Arbitrum, with off-chain order-book matching and AMM-fallback liquidity. The distinguishing feature is the integrated multi-product margin account — spot + perp + money-market positions share a single margin pool, enabling cross-product netting. Settlement asset: USDC. Max leverage: 10x on majors, lower on alts. Asset coverage: 30-50 perp pairs + spot. MEV protection: medium-high (off-chain matching). Gas economics: Arbitrum L2 gas (sub-cent). Fee tier: maker 0.0% / taker 0.02-0.04%. Daily-volume range: $0.1-0.4B. Market share: top-10 perp DEX.

Vertex Edge (multi-chain expansion) extends the order book across Arbitrum, Mantle, Sei, Base, Blast, and other chains via a synced-orderbook architecture, attempting to capture liquidity across multiple L2s without forcing traders onto a single chain. Detail: cross-link to AMM design evolution for the hybrid CLOB+AMM lineage.

Drift Protocol (Solana, 2021-)

Solana-native perp DEX using vAMM (virtual AMM) + oracle pricing + JIT auction model. Drift v2 (2022) replaced the earlier purely-vAMM design with a hybrid where takers can be filled by JIT (just-in-time) market-maker bots responding within a short auction window before the vAMM fallback. Settlement asset: USDC. Max leverage: 20x on majors, 10x on alts. Asset coverage: 40+ perp pairs. MEV protection: medium (Solana’s block-leader architecture has known MEV vectors, mitigated by Drift’s JIT auction). Gas economics: Solana network fees (sub-cent). Fee tier: maker 0.01% / taker 0.10% standard with tier discounts. Daily-volume range: $0.1-0.3B. Market share: top Solana perp DEX, top-10 globally.

Drift’s significance is being the central Solana-ecosystem perp hub, benefiting from Solana’s general resurgence through 2024-2025 and from being the default integration target for Solana-native applications adding perp exposure. DRIFT token launched 2024. Detail: cross-link to Solana ecosystem DEX comparison for the Solana-specific context.

GMX v1 (Arbitrum + Avalanche, 2021-)

The prototypical pool-counterparty perp DEX — a single multi-asset pool (GLP, holding BTC/ETH/USDC/stables) serves as the counterparty to every trader. Trader PnL is the GLP holder’s loss/gain. Prices come from Chainlink oracles with internal keepers. Settlement asset: pool assets (BTC/ETH/USDC). Max leverage: 50x on majors. Asset coverage: limited to GLP-pool-supported assets (BTC, ETH, LINK, UNI + a few). MEV protection: low-medium (oracle-update timing creates known front-running surfaces, addressed in v2). Gas economics: Arbitrum or Avalanche L1/L2 gas (low). Fee tier: 0.1% open/close + dynamic borrow rate. Daily-volume range: $0.05-0.2B (down from $0.5-1B peak in 2022-2023). Market share: was 2022-2023 perp-DEX leader; now mid-tier.

GMX v1 invented the “GLP as universal counterparty” pattern that many forks (Gains Network, Vela, others) copied. The structural challenge is toxic flow — sophisticated traders can systematically extract value from the GLP pool when the oracle update lags real market prices.

GMX v2 (Arbitrum + Avalanche, 2023-)

Architectural rebuild responding to v1’s toxic-flow problem. Introduces per-market isolated pools (GM pools) instead of the single GLP, funding rates to balance long/short skew, and improved oracle-pricing mechanics. Settlement asset: per-market backing asset. Max leverage: 100x on majors (specific markets). Asset coverage: expanding via per-market pools. MEV protection: medium (improved over v1). Gas economics: Arbitrum gas (low). Fee tier: 0.04-0.07% open/close + funding. Daily-volume range: $0.1-0.3B. Market share: top-10 perp DEX.

v2 is GMX’s attempt to remain relevant after Hyperliquid and dYdX v4 took share. The per-market pool model trades the simplicity of GLP for better risk-isolation, at the cost of fragmenting LP capital across markets.

Synthetix Perps (Optimism + Base, 2022-)

Perps built on the Synthetix synth-debt-pool model — the SNX-staker debt pool provides counterparty liquidity, oracle pricing (Pyth) gives execution price, and traders take leveraged positions against the pool. Perps v2 (2023) added improved oracle integration; Perps v3 (2024) modularised the architecture allowing front-ends to plug into the Synthetix backend. Settlement asset: sUSD. Max leverage: 50x on majors. Asset coverage: 40+ pairs (limited by oracle availability). MEV protection: medium (oracle-based pricing has timing-attack surfaces, addressed by Pyth pull-oracle mechanism). Gas economics: Optimism / Base L2 gas (low). Fee tier: maker 0.02% / taker 0.05-0.10% by market. Daily-volume range: $0.05-0.2B (much routed via Kwenta, see below). Market share: top-15.

Synthetix’s significance is being the backend infrastructure that multiple front-ends (Kwenta, Polynomial, Lyra-adjacent products) build on, rather than competing as a consumer-facing DEX directly.

Kwenta (Optimism + Base, 2022-)

Front-end DEX built on Synthetix Perps backend, designed for active traders. KWENTA token holders share fees with stakers. Settlement asset: sUSD (inherited from Synthetix). Max leverage: 50x. Asset coverage: same as Synthetix Perps (40+). MEV protection: medium. Gas economics: Optimism / Base gas. Fee tier: same as Synthetix backend + Kwenta front-end fee. Daily-volume range: $0.02-0.1B. Market share: top-20.

Kwenta’s existence illustrates the front-end / backend separation that Synthetix v3 architected — the consumer brand is decoupled from the liquidity backend, which lets multiple front-ends compete on UX while sharing a single liquidity pool.

Aark Digital (Arbitrum, 2023-)

LP-pool model perp DEX with a vault-shareholder counterparty design similar to GMX but with different risk-management and a focus on capital-efficient market-making. Settlement asset: USDC. Max leverage: 20x on majors. Asset coverage: 20-30 pairs. MEV protection: medium. Gas economics: Arbitrum gas. Fee tier: 0.05-0.10%. Daily-volume range: $0.01-0.05B. Market share: top-30 niche.

Aark is a second-wave entrant competing in the pool-counterparty segment where GMX, GNS, and similar protocols already established the pattern. Differentiation is on the LP risk-management mechanics and capital efficiency.

Aevo (OP Stack appchain, 2024-)

Successor to Ribbon Finance’s options product, Aevo runs a CLOB-based options + perp venue on a custom OP Stack appchain. Perps share infrastructure with the options book, enabling cross-product margin. Settlement asset: USDC. Max leverage: 20x on majors. Asset coverage: 40+ pairs (perps + options). MEV protection: high (off-chain matching with on-chain settlement). Gas economics: OP Stack L2 gas (low). Fee tier: maker 0.03% / taker 0.05%. Daily-volume range: $0.05-0.2B. Market share: top-15 perp + dominant on-chain options.

Aevo’s significance is being the options-and-perps integrated venue rather than a pure-perp DEX, which serves a different trader segment (volatility / hedging-oriented flows alongside leveraged directional).

Perp DEX using StarkEx (similar architectural base to old dYdX v3) and zkLink extensions for cross-chain liquidity. Settlement asset: USDC. Max leverage: 50x. Asset coverage: 50+ pairs. MEV protection: medium-high (off-chain matching). Gas economics: trader pays no gas (matching off-chain); deposit/withdraw uses underlying chain gas. Fee tier: maker 0.02% / taker 0.05%. Daily-volume range: $0.05-0.2B. Market share: top-15, with strong Asia retail presence.

Apex’s differentiation is strong Asia retail distribution (significant non-English-language community presence and CEX-style UX) competing for the trader segment that dYdX v3 served before its Cosmos migration.

Big comparison matrix table

DEXChain / baseLiquidity modelSettlement assetMax leverageAsset coverageFee tier (maker/taker)MEV protectionGas economics for traderDaily volume range (USD)Perp-DEX market share band
dYdX v3StarkEx L2 (Ethereum)CLOB (off-chain match)USDC20x~40 pairs0.02% / 0.05%HighNone (off-chain)$0.5-2B (2022-2023)was #1 in 2022
dYdX v4Cosmos appchain (dYdX Chain)CLOB (validator-matched, fully on-chain)USDC20x40+0.02% / 0.05%HighValidator-network gas (cents)$0.3-1Btop-3
HyperliquidHyperliquid L1 (HyperBFT + HyperEVM)CLOB (fully on-chain)USDC (bridged Arb)50x100+rebate / 0.025-0.05%HighHyperEVM gas (sub-cent)$1-5B#1 by volume
VertexArbitrumCLOB + AMM hybridUSDC10x30-500.0% / 0.02-0.04%Medium-highArbitrum gas (sub-cent)$0.1-0.4Btop-10
DriftSolanavAMM + oracle + JIT auctionUSDC20x40+0.01% / 0.10%MediumSolana fees (sub-cent)$0.1-0.3Btop-10 / Solana #1
GMX v1Arbitrum + AvalancheGLP pool-counterparty + oraclepool assets50xlimited (BTC/ETH/LINK/UNI+)0.1% + borrowLow-mediumArb/Avax gas$0.05-0.2Bmid-tier
GMX v2Arbitrum + AvalanchePer-market GM pools + fundingper-market100xexpanding0.04-0.07% + fundingMediumArb gas (low)$0.1-0.3Btop-10
Synthetix PerpsOptimism + BaseSynth-debt-pool + Pyth oraclesUSD50x40+0.02% / 0.05-0.10%MediumOP / Base gas (low)$0.05-0.2Btop-15 backend
KwentaOptimism + BaseSynthetix backend (front-end DEX)sUSD50x40+same + front-end feeMediumOP / Base gas$0.02-0.1Btop-20 front-end
Aark DigitalArbitrumLP-vault counterpartyUSDC20x20-300.05-0.10%MediumArb gas$0.01-0.05Btop-30 niche
AevoOP Stack appchainCLOB (options + perps shared book)USDC20x40+ (perps + options)0.03% / 0.05%HighOP Stack gas (low)$0.05-0.2Btop-15 + options leader
Apex ProtocolStarkEx + zkLinkCLOB (off-chain match)USDC50x50+0.02% / 0.05%Medium-highNone (off-chain match)$0.05-0.2Btop-15, Asia retail strong

Architecture cross-axis: CLOB vs vAMM vs hybrid

The matrix above lays out 11 venues but they collapse into three architectural families that determine fee profile, MEV resistance, and LP economics:

CLOB family (fully on-chain or hybrid match)

  • Hyperliquid — fully on-chain CLOB on custom L1
  • dYdX v4 — fully on-chain CLOB on Cosmos appchain (validator-matched)
  • dYdX v3 — off-chain match + on-chain settlement (StarkEx)
  • Apex — off-chain match + on-chain settlement (StarkEx)
  • Aevo — off-chain match + on-chain settlement (OP Stack appchain)
  • Vertex — hybrid CLOB + AMM fallback

Trader who fits CLOB: high-frequency / professional traders, those who need limit / stop / IOC / FOK order types, those sensitive to slippage on size, those needing tight bid-ask spreads on majors. Fee economics are typically maker rebates + taker fees (familiar to CEX-arbitrage traders). The CLOB family captured share aggressively through 2024-2026 with Hyperliquid leading.

vAMM / pool-counterparty family

  • GMX v1 — GLP single multi-asset pool
  • GMX v2 — per-market GM pools
  • Aark Digital — LP-vault counterparty (GMX-lineage)
  • Drift — vAMM + JIT auction (Solana)
  • Synthetix Perps — synth-debt-pool

Trader who fits vAMM: traders comfortable with oracle-pricing slippage models, those who prefer “everyone trades at oracle mid” semantics, smaller-size traders for whom CLOB depth is not the binding constraint. LP-side economics dominate — the LP / debt-pool holder is the structural counterparty to all trader PnL, which creates a different risk-return profile than passive LP’ing in spot AMMs. The model has structural toxic-flow vulnerability that v2 generations (GMX v2, Drift v2) work hard to mitigate.

Front-end on shared backend family

  • Kwenta — front-end on Synthetix Perps backend

Significance: the front-end / backend separation is what Synthetix v3 architected as a deliberate platform play. Other protocols (Polynomial, Lyra-adjacent) follow the same model. The front-end-only DEX competes on UX, fee-share, and trader segment targeting, while sharing the same underlying liquidity pool.

Trader-segment fit

Beyond architecture, perp DEXs differentiate on trader-segment fit:

Trader segmentBest-fit venuesWhy
Professional / HF arbitrageHyperliquid, dYdX v4CLOB depth, low latency, MEV protection
High-leverage degensHyperliquid (50x), GMX v2 (100x), Apex (50x), Drift (20x)Leverage ceiling matters
Long-tail asset tradersHyperliquid (100+ pairs), Drift, AevoAsset coverage breadth
LST / restaking collateral(chain-specific) — Drift accepts JitoSOL; some EVM venues accept stETHLST-as-margin economics
Solana-nativeDriftSolana ecosystem depth
Options + perps cross-marginAevoIntegrated options book
Multi-product (spot + perp + money market)VertexUnified margin account
Asia retail (non-English-first UX)Apex, HyperliquidDistribution + UX choices
Institutional / DeFi-native fundsdYdX v4, HyperliquidDecentralisation + auditability

The LST / restaking collateral row connects to liquid-staking / restaking CEX exposure for the LST-as-collateral mechanics across venues.

Boundary cases

  • CEX-equivalent or not? — Hyperliquid’s order-book latency and depth on top pairs approach CEX comparability (Binance perp, Bybit perp). However, the regulatory posture differs entirely: Hyperliquid geo-blocks US traders, runs no centralised KYC, and operates with no jurisdictional licence. A CEX-comparable UX should not be confused with CEX-equivalent regulatory standing.
  • dYdX v3 vs v4 read as one product or two? — The architectural rebuild was extensive enough (different chain, different consensus, different token utility) that they are best read as two products from the same team rather than one product across versions. The matrix gives each its own row.
  • GMX v1 vs v2 read as one product or two? — Similar to dYdX, GMX v1 (GLP single-pool) and v2 (per-market GM pools + funding) are structurally different enough to warrant separate rows. Both remain live as of 2026 with different LP profiles.
  • Synthetix Perps and Kwenta double-counting — Kwenta is a front-end on Synthetix Perps’ backend. Daily-volume figures for Kwenta are largely also counted in Synthetix Perps’ volume, since Kwenta routes orders to the Synthetix backend. Cross-protocol volume reporting (e.g., DefiLlama derivatives) sometimes double-counts; the matrix separates the rows for clarity but a reader should aggregate them when computing total Synthetix-stack flow.
  • Drift on Solana vs the rest on EVM — Solana’s account-model and block-leader architecture changes MEV dynamics, slot-time guarantees, and gas economics in ways that don’t map cleanly to EVM analogies. Comparison of Drift’s MEV-protection vs EVM peers requires reading the Solana-specific JIT-auction model rather than treating MEV protection as a single dimension.
  • Aevo as “options + perps” rather than “perp DEX” — Aevo’s economic identity is the integrated options-and-perp venue; treating it purely as a perp DEX understates the options-side flow that uses the same margin account. The matrix includes Aevo but flags that the options book is its primary differentiator.
  • Hyperliquid HYPE token TGE economics — Hyperliquid’s 2024-11 token launch with no pre-sale and emission distributed to retroactive users is structurally different from dYdX, GMX, Drift, and Synthetix token economics. Token-side comparisons require separate analysis (not covered in this matrix’s volume / fee axes).
  • Cross-chain settlement on/off-ramp — Several perp DEXs use USDC as settlement asset but the on-ramp routes differ (Arbitrum-native, Solana-native, Cosmos appchain, OP Stack). The settlement-asset row in the matrix records the canonical asset, but the practical on-ramp UX is a separate dimension — see cross-chain bridge × CEX deposit/withdrawal for the bridge-side detail.
  • Volume figures are approximate ranges, not point estimates — Perp DEX daily volumes swing 5-10x with market volatility. The “daily volume range” column gives an active-market reasonable range based on 2024-2026 observation; readers should consult DefiLlama derivatives for current point-in-time figures.
  • JP and US regulatory exposure — None of the perp DEXs in this matrix appear on JVCEA WhiteList for JP-side native-token listing, and most geo-block US traders following CFTC / SEC engagement. Reading this matrix against JP CEX equivalents like the JP CEX parent FG adjacency matrix surfaces the regulatory-licensing gap between on-chain perp DEX and JP-domestic VASP.

Settlement-asset side-by-side

Settlement assetUsed byImplication
USDCdYdX v3, dYdX v4, Hyperliquid, Vertex, Drift, GMX (USDC slice of GLP), Aark, Aevo, ApexDefault modern stablecoin; on-chain availability of USDC across L2s and Solana is sufficient; Circle counterparty / issuance risk is shared across most of the matrix
sUSDSynthetix Perps, KwentaSynthetix-native algorithmic stable backed by SNX debt-pool; different risk profile from USDC — not Circle-issued
pool assets (BTC / ETH / LINK / UNI / USDC)GMX v1GLP pool composition is the de-facto settlement basket; multi-asset risk exposure at LP level
per-market backing assetGMX v2Per-market GM pool composition; risk isolated by market

The USDC-dominated settlement landscape is structurally significant — it concentrates Circle issuance / regulatory / reserve-management risk as a cross-DEX systemic factor. If Circle were to face issuance disruption, the majority of the perp-DEX volume in the matrix would be affected simultaneously. This is one of the under-acknowledged systemic risk axes in on-chain derivatives.

Chain-base footprint summary

Chain / basePerp DEXs hostedCumulative perp-DEX share contribution
Hyperliquid L1Hyperliquidvery high (single venue, dominant market share)
Cosmos appchain (dYdX Chain)dYdX v4high
ArbitrumVertex, GMX v1/v2, Aarkmedium-high (multi-venue)
AvalancheGMX v1/v2low-medium
SolanaDriftmedium (single dominant venue on chain)
OptimismSynthetix Perps, Kwentalow-medium
BaseSynthetix Perps (post-Base launch), Kwentalow-medium (growing)
OP Stack appchainAevolow-medium
StarkEx L2dYdX v3 (legacy), Apexlow (historical + Asia-retail niche)
zkLinkApex (cross-chain extension)very low

The chain-base distribution shows purpose-built infrastructure (Hyperliquid L1, dYdX Cosmos appchain, OP Stack appchains) winning share over shared-L2 infrastructure (Arbitrum, Optimism, Base) through 2024-2025. The chain-native flip pattern documented in native DEX flip-incumbent pattern applies — venues that control their own chain capture the volume that chain-tenant venues forfeit to chain congestion and shared-MEV exposure.

Architectural-evolution timeline

YearEventArchitectural significance
2018-08dYdX margin trading (precursor) launchFirst non-derivatives DEX experiment with leverage
2020Perpetual Protocol v1 (vAMM on xDai)First vAMM perp at scale (predecessor to GMX-lineage)
2021-04dYdX v3 launch on StarkEx L2First CLOB perp DEX at scale; CEX-comparable UX
2021-08GMX launch on ArbitrumGLP pool-counterparty pattern established
2021-09Drift v1 launch on SolanaSolana-native perp DEX (purely vAMM initially)
2022Synthetix Perps v1 + Kwenta launchSynth-debt-pool perp model, front-end/backend split begins
2023-06dYdX v4 mainnet on CosmosFirst fully-on-chain CLOB perp DEX
2023-08GMX v2 launchPer-market GM pools + funding model
2023-Q3Vertex Protocol launchHybrid CLOB+AMM with multi-product margin account
2024-Q1Hyperliquid mainnet maturationCustom L1 fully-on-chain CLOB takes share
2024-Q2Aevo launch on OP Stack appchainOptions + perps integrated CLOB venue
2024-11Hyperliquid HYPE TGENo-pre-sale emission-to-retroactive-users distribution
2025Cross-chain perp aggregator emergenceVertex Edge, intent-based perp routing experiments

The trajectory shows CLOB winning architecturally through 2024-2025 (Hyperliquid + dYdX v4 leading on volume), with pool-counterparty / vAMM models retaining a niche in specific positioning (Drift on Solana, GMX v2 for high-leverage isolated markets). The next architectural inflection is plausibly intent-based cross-chain perp routing (allowing a trader to execute on whichever chain has best liquidity without bridge UX), but that pattern is not yet at scale as of 2026.

Fee-economics summary

A practical fee-take-rate comparison (taker side, before rebates and tier discounts):

DEXStandard taker feeFunding mechanismEffective annualized cost for 1x position
Hyperliquid0.025-0.05%Funding rate floats with skewdepends on funding (volatile)
dYdX v40.05%Funding rate floatsdepends on funding
dYdX v30.05%Funding rate floatsdepends on funding
Vertex0.02-0.04%Funding ratelow fee + funding
Drift0.10%Funding ratemedium fee + funding
GMX v10.10% open + 0.10% closeBorrow rate (continuous, not 0-sum)higher cost — borrow is structurally positive
GMX v20.04-0.07% open + closeFunding ratemedium
Synthetix Perps0.05-0.10% (market dependent)Funding ratemedium
Kwentasame + front-endFunding ratemedium-high
Aark0.05-0.10%Fundingmedium
Aevo0.05%Fundinglow-medium
Apex0.05%Fundinglow-medium

Key economic distinction: GMX v1’s borrow-rate model is structurally cost-positive for the trader (the LP earns continuous borrow), unlike funding-rate models which are zero-sum between longs and shorts. This makes GMX v1 less attractive for hold-long-duration positions and more attractive for quick directional bets, while funding-rate venues can be cheaper for long-hold positions depending on funding direction. Understanding the fee + funding/borrow stack is essential for cross-venue comparison.

Sources