# Crypto Derivatives Exchanges Compared on One Sheet Updated: 24 September 2026 | Book readings 22 September 2026 | The Margin Brief desk ## Summary This page holds crypto derivatives exchanges compared parameter by parameter on one sheet, in the site's standing score order, led by Lighter at 12.5 of 15. The columns disagree on purpose: the cheapest venue is not the one with the widest book, and no column on this sheet settles the question on its own. Weighting: Exit liquidity 30 · Collateral and settlement 22 · Cost per round trip 20 · Loss record and audits 14 · Contracts and asset classes 9 · Getting a position on 5, fixed 22 September 2026. Method: https://insurancerefocused.com/how-we-rate ## What the figures decide 1. Widest book measured: edgeX at $29.3 million; the thinnest, Paradex at $55.00, could not fill a $1 million order. 2. Nothing to pay at the entry tier: Lighter and Paradex; dearest is Aevo, $160.00 on $100,000 of exposure. 3. Longest contract list: Aster at 578; the shortest here is EVEDEX at 52. 4. The two venues documenting an operator-free withdrawal, Lighter and edgeX, are not the four that publish a bug bounty. ## Eight venues in the site's standing order, September 2026 The order is the site-wide total, not a reading made for this page: a sheet that reshuffled itself would be arguing a second verdict. | # | Venue | Score of 15 | Maker / taker | Markets | Model | |---|---|---|---|---|---| | 1 | Lighter | 12.5 | 0% / 0% | 214 | Hybrid, ZK-proven | | 2= | edgeX | 11.5 | 0.040% / 0.045% | 172 | Hybrid | | 2= | Extended | 11.5 | 0% / 0.025% | 325 | Hybrid | | 4 | Aster | 11.3 | 0% / 0.040% | 578 | Own layer 1, operator-run | | 5 | Hyperliquid | 10.2 | 0.015% / 0.045% | 324 | On-chain order book | | 6 | EVEDEX | 9.7 | 0.015% / 0.045% | 52 | Hybrid | | 7 | Paradex | 9.0 | 0% / 0% | 63 | Hybrid, Starknet appchain | | 8 | Aevo | 7.7 | 0.050% / 0.080% | 102 | Hybrid | Venue facts as of 18 September 2026; entry-tier fees on the BTC perpetual. Access differs by country. ## Assessments ### 1. [Lighter](https://lighter.xyz) — the highest total, on a schedule of nothing Score 12.5/15. Lighter's default account pays no maker or taker fee, by its documentation of 18 September 2026. Its BTC book held $16.9 million, third widest here. - Works: Nine audit reports on file, level with edgeX and more than any other venue here; A priority exit queue on Ethereum, usable without the operator. - Falls short: The free account adds 300 ms to a taker order; No published bug bounty. - Not for: traders who want a bounty behind the code. ### 2. [edgeX](https://pro.edgex.exchange) — the widest depth, at the second dearest maker fee Score 11.5/15. edgeX matches off the chain and writes balances and trades to a chain of its own, by its documentation of 18 September 2026. A $1 million order met 0.22 bps, the least of the eight. - Works: Median BTC depth of $29.3 million, the widest here; An on-chain withdrawal that bypasses the operator. - Falls short: 0.040% maker, dearer at the entry tier than every venue here but Aevo; No bug bounty programme in its documentation or help centre, read 18 September 2026. - Not for: makers who trade often. ### 2. [Extended](https://extended.exchange/) — full marks on security inputs, on a thin book Score 11.5/15. Extended reports 325 perpetual markets across six asset classes, settled on Starknet ([Extended docs](https://docs.extended.exchange/extended-resources/trading/rwa-markets), checked 18 September 2026). - Works: Bug bounty of up to $500,000, and no recorded loss of user funds; No maker fee on any account, not only a top tier. - Falls short: 278 of its 325 markets are quote-driven, with no public book; Median BTC depth of $7.8 million, thinnest of the top four. - Not for: traders who want an operator-free exit. ### 4. [Aster](https://www.asterdex.com) — the longest contract list, on its own chain Score 11.3/15. Aster lists 578 perpetual markets on its own layer 1 ([Aster docs](https://docs.asterdex.com/overview/what-is-aster.md), checked 18 September 2026). Makers have paid nothing since 2 February 2026; takers pay 0.040%. Its spread, 0.01 bps, was the tightest we read, level with Lighter. - Works: 578 perpetual markets, the longest list here; Immunefi bug bounty of up to $200,000, live since April 2025. - Falls short: Closed core code, with no outside validators yet; The 200x bracket applies only to the first 400 USDT of notional. - Not for: traders who want open validators. ### 5. [Hyperliquid](https://hyperliquid.xyz) — the largest reported open interest, the lowest BTC cap Score 10.2/15. Hyperliquid keeps the order book on its own chain ([Hyperliquid docs](https://hyperliquid.gitbook.io/hyperliquid-docs/about-hyperliquid.md), checked 18 September 2026). CoinMarketCap put its open interest at $8.7 billion on 22 September 2026, the largest of the eight. - Works: Every order, trade and liquidation recorded on its chain; Bug bounty of up to 1 million USDC. - Falls short: About $4 million left the vault it runs in March 2025, and roughly $4.9 million in November 2025; No documented exit once the operator stops. - Not for: traders wary of a venue-run vault. ### 6. EVEDEX — the shortest contract list, across six asset classes Score 9.7/15. The EVEDEX list runs to 52 perpetual contracts over six asset classes: 39 crypto markets, five US stocks, an index, gold through XAUT, silver, crude oil, a pair of FX markets and two pre-IPO contracts, all stated to trade around the clock (EVEDEX documentation, our reading of 18 September 2026). One cross-margined balance in USDT carries them, at 0.015% maker / 0.045% taker before cashback. - Works: Median BTC depth of $18.4 million, second widest here; Six asset classes on one cross-margined balance. - Falls short: Position data reaches Arbitrum in batches rather than trade by trade; No bug bounty programme listed on CertiK Skynet; 52 contracts against 578 here, with no spot and no options. - Not for: UK retail clients, and traders needing an operator-free exit. ### 7. [Paradex](https://www.paradex.trade) — nothing to pay, and almost nothing resting Score 9.0/15. Paradex charges retail orders nothing and settles on a Starknet-stack appchain, by its documentation of 18 September 2026. A $100,000 order moved the mid 162.91 bps, and $1 million found no fill at all. - Works: Sherlock bug bounty of up to 500,000 USDC; No maker or taker fee on retail orders. - Falls short: Median resting depth of $55.00, the thinnest here; A lockdown procedure lets the operator limit withdrawals. - Not for: anyone closing size at market. ### 8. [Aevo](https://www.aevo.xyz/) — the dearest round trip, and a loss on the record Score 7.7/15. Aevo matches off-chain and settles on its own rollup, batched to Ethereum hourly, by its documentation of 18 September 2026. It lists 102 perpetual contracts at 0.050% maker / 0.080% taker, $160.00 on $100,000 of exposure. - Works: 102 contracts across five classes beyond crypto; Every matched trade settles in contracts on its rollup. - Falls short: 0.080% taker, the dearest round trip here; About $2.7 million drained from its legacy vaults in December 2025, a 19% haircut. - Not for: cost-sensitive takers. ## Sources quoted > "Currently, Extended features a flat fee structure for both perpetuals and spot markets: Taker: 0.025% Maker: 0.000%" — Extended documentation, Trading fees and rebates, 18 September 2026. https://docs.extended.exchange/extended-resources/trading/trading-fees-and-rebates > "HyperCore includes fully onchain perpetual futures and spot order books." — Hyperliquid documentation, About Hyperliquid, 18 September 2026. https://hyperliquid.gitbook.io/hyperliquid-docs/about-hyperliquid.md > "Orders are only stored on the Paradex Cloud and are only known to the operators of Paradex Cloud." — Paradex documentation, Trading privacy, 18 September 2026. https://docs.paradex.trade/trading/privacy.md ## Crypto derivatives exchanges compared, parameter by parameter Venue facts come from each venue's own documents, read on 18 September 2026; depth, spread and impact come from the reading. All eight margin across the account rather than per position. Posted collateral is the whole of what one of these contracts can lose. | Parameter | Lighter | edgeX | Extended | Aster | Hyperliquid | EVEDEX | Paradex | Aevo | |---|---|---|---|---|---|---|---|---| | Median depth, 10 bps | $16.9 million | $29.3 million | $7.8 million | $14.8 million | $7.4 million | $18.4 million | $55.00 | $1.4 million | | Median spread | 0.01 bps | 0.04 bps | 0.12 bps | 0.01 bps | 0.12 bps | 4.12 bps | 18.39 bps | 2.43 bps | | Impact, $1 million order | 1.48 bps | 0.22 bps | 0.70 bps | 1.11 bps | 0.72 bps | 2.75 bps | not filled | 7.63 bps | | BTC cap and notional | 50x, 2% margin | 100x to $1 million | 50x to $4 million | 200x to 400 USDT | 40x to $150 million | 200x to $50,000 | 50x, 300 BTC limit | 20x, 5% margin | | Settlement | Batch proofs, Ethereum | Every trade, own chain | Every transaction, Starknet | Every trade, own layer 1 | Every order and trade, own chain | Position data to Arbitrum in batches | Every trade, own appchain | Every trade, hourly batches | | Operator-free exit | Documented | Documented | Not documented | Not documented | Not documented | Not documented | Not documented | Not documented | | Audit reports on file | 9 | 6 | 2 | 7 | 2 | 2 | 5 | 5 | | Bug bounty | None | None | Up to $500,000 | Immunefi, $200,000 | Up to 1 million USDC | None | Sherlock, 500,000 USDC | None | | Funds lost in 24 months | None | None | None | None | About $4.9 million and $4 million, vault | None | None | About $2.7 million, legacy vaults | | Asset classes beyond crypto | 5 | 3 | 5 | 5 | 4 | 5 | 3 | 5 | A loss means money not repaid: a compensated outage and a market-driven liquidation are not counted, and scoring credits two audit reports at most. ## Where the sheet disagrees with itself The cheapest entry tier belongs to Lighter and Paradex, which charge retail orders nothing; the widest resting depth to edgeX, whose 0.040% maker sits second on this sheet behind Aevo's 0.050%. Hyperliquid reported the largest open interest, $8.7 billion on CoinMarketCap on 22 September 2026, on a book holding $7.4 million — behind five venues here. A sheet has no opinion about which column matters; the six [weights](/how-we-rate) do, and they were fixed on 22 September 2026, before any venue was scored. ## How the figures were taken 8 venues, one set of six weights closed on 22 September 2026 ([how we rate](/how-we-rate)). Schedules, rungs, markets and audits come from each venue's own documents, read on 18 September 2026. Nobody publishes an exit column: depth, spread and impact come from a sampling of every public BTC perpetual book over 122 polls between 21 September 2026 and 22 September 2026, without an order. The readings are of the BTC perpetual only, and a smaller market is thinner. Entry-tier rates ignore cashback and volume tiers. Settlement and security rest on what each venue publishes, which is not complete. ## FAQ ### How do you compare crypto derivatives exchanges? Parameter by parameter, and then with a published weighting. This sheet sets the same parameters side by side for eight venues on 24 September 2026: three readings taken off the public books, the margin ladder, settlement, audits, bounties, the loss record and asset classes. ### Do more perpetual markets make an exchange better? Not on any reading here. Aster lists 578 perpetual contracts and Extended 325, and neither held the widest book: edgeX did, with a median $29.3 million resting within 10 basis points. A longer list widens what can be traded, not what can be closed. ### What does on-chain settlement mean on a hybrid exchange? That matching happens on the venue's own servers while the record of balances and trades is written to a chain. The cadence differs: Extended validates every transaction on Starknet, Aevo batches hourly, and EVEDEX writes position data once enough has changed, by its documentation of 18 September 2026. ### Which crypto derivatives exchanges run a bug bounty? Four of the eight, at the figures each publishes: Hyperliquid up to 1 million USDC, Paradex up to 500,000 USDC through Sherlock, Extended up to $500,000 and Aster up to $200,000 through Immunefi. The other four, Lighter, edgeX, EVEDEX and Aevo, publish none, read 18 September 2026. ### Have any perpetual exchanges lost user funds in the last two years? Two of the eight. A venue-run vault at Hyperliquid lost about $4.9 million in November 2025 and about $4 million in March 2025. Aevo's legacy vaults were drained of about $2.7 million in December 2025, a 19% haircut for holders. A liquidation from a moving market is not counted. ### How many security audits does a derivatives exchange usually have? There is no usual number. On this sheet the count runs from two, at Extended, Hyperliquid and EVEDEX, to nine at Lighter, with edgeX on six and Aster on seven. Our scoring counts two at most: a third adds little the first two did not. ### Is a hybrid exchange a DEX or a CEX? Neither label fits. Six of the eight here match orders on servers the operator runs and write the result to a chain. Hyperliquid keeps the book itself on a chain; Aster runs its own layer 1, with closed core code and no outside validators yet. ### What happens to an open position if an exchange stops matching orders? It stays open, and so does the margin behind it. Two venues document a way out that does not need the operator: Lighter through a priority queue on Ethereum, edgeX on-chain. The other six describe no such route in the documents read on 18 September 2026. ### Why do two exchanges with the same fee cost different amounts to trade? Because the fee is charged on notional and the exit is charged by the book. Hyperliquid and EVEDEX both publish 0.015% maker / 0.045% taker. On $1 million of bitcoin the measured impact was 0.72 bps at one and 2.75 bps at the other. ### Can one exchange be ahead on every parameter? Not among these eight. Lighter holds the highest total at 12.5 of 15 and charges nothing, yet its book was third widest and it publishes no bug bounty. edgeX held the widest book, and only Aevo charges a dearer maker. Extended runs a bounty of up to $500,000 on the thinnest book of the scored four. ## Editorial note These are leveraged perpetual futures: the margin behind a position is all of what it can lose. Placement here is paid for; the order of the table is not. Corrections: editorial@insurancerefocused.com. The Margin Brief desk, 24 September 2026 Placement in this file is paid for; the order of every table follows the published formula and nothing else.