Best DEX for Scalping in 2026: Zero-Fee and Fastest Perp DEXs Ranked

The best DEX for scalping is Hyperliquid for depth and Lighter for zero fees. Lighter’s default Standard account charges 0% maker and 0% taker on every perp market, at the price of a 300 ms delay on taker orders. Hyperliquid has the deepest DEX order books ($4–8B of perp volume a day), 0.045% taker and 0.015% maker, 4% less with code HOLYGRAIL. For maker-only scalping, Aster charges 0% maker at every tier. Fees and limits come from each exchange’s official docs (verified September 25–27, 2026); order-book and hourly data from their public APIs.

1. The 5 best DEXs for scalping, ranked

1. Hyperliquid — Best overall for scalping. Deepest DEX books ($4–8B/day, 178 perps as of October 20, 2026), cancels processed before takers in every block, 0.015% maker via Chase and post-only orders, 0.045% taker (4% less with code HOLYGRAIL).
2. Lighter — Best zero-fee scalping DEX. Standard accounts pay 0% maker and taker on every market; taker orders wait 300 ms. Premium: 140 ms for 0.028% taker.
3. Aster — Best for maker-only scalping. 0% maker at every VIP level, a native Chase order that stays post-only, 0.04% taker.
4. edgeX — Fast engine, costly maker side. Claims “microsecond latency” for its matching engine (edgeX V2 docs; no end-to-end figure published), but charges 0.04% maker next to 0.045% taker, so post-only scalping saves little.
5. dYdX — Established, slower. 0.05% taker / 0.01% maker at tier 1; its docs put order-to-trade at about one block (~800 ms).

The ranking weighs what a scalper pays on every trade (round-trip fees, especially on the maker side), how quickly orders fill and cancel, and how much size the book absorbs near the mid price. Volume and market count come next. Pick by how you trade:

2. What makes a DEX good for scalping (and why AMMs are not)

Four things decide whether a venue works for scalping: a central limit order book (so you can post limit orders and choose to be the maker), low round-trip fees, ideally with a cheap or free maker side, fast fills and cancels, and tight spreads with depth on the markets you trade. Every venue ranked above is an order-book perp DEX.

Swap DEXs built on automated market makers (AMMs) fail almost every test. On Uniswap you pay the pool fee on each swap (0.05% or 0.3% in the common v3 tiers), plus gas and price impact, and you wait for a block to confirm. A public-mempool swap can also be sandwiched by MEV bots (see sandwich attack protection). There is no maker side to earn and no resting stop-loss. Pool-based perps are better but still costly for scalping: GMX charges 0.04%–0.06% to open and again to close, and Jupiter Perps 0.06% each way, plus borrow fees. The full comparison is in order book vs AMM, and the wider list of spot and perp venues is on our best DEX page.

3. Why fees decide scalping: the break-even math

A scalp holds for seconds or minutes and targets a small move, often 0.1–0.5%. Every trade pays a round trip — one fee to open, one to close — and on a taker-only venue that is two taker fees. Take a scalp that aims for +0.4% and stops out at −0.2% (a 2:1 reward-to-risk). With round-trip cost c, a winner nets 0.4% − c and a loser costs 0.2% + c, so the break-even win rate is (0.2% + c) ÷ 0.6%:

• Zero fees (Lighter Standard, or maker-only on Aster): 33.3%.
• Aster taker (0.08% round trip): 46.7%.
• Hyperliquid taker (0.09%): 48.3%; 47.7% with code HOLYGRAIL.
• Binance USDⓈ-M taker, regular tier (0.1%): 50.0%.

A 15-point gap in the win rate you need is, for many scalping systems, the entire edge. It is also the honest answer to “can scalping be profitable?”: only if your system wins more often than the break-even line your fees draw. The same logic explains why maker orders matter so much: a Hyperliquid scalper who enters and exits with post-only orders pays 0.0288% per round trip with the code and needs about 38.1%.

4. Cost per round trip: DEXs vs Binance, Bybit and OKX

Fee schedules from the Hyperliquid, Lighter, Aster, edgeX, dYdX and Paradex docs and the Binance, Bybit and OKX fee pages, verified September 25–27, 2026 (facts registry). Hyperliquid “with code” = 4% referral discount on the first $25M of volume. Regular or entry tiers, no token discounts (BNB, MNT, HYPE, LIT) assumed. Slippage and funding excluded.

Three lessons. On the taker side Lighter Standard is unbeatable and Lighter Premium (0.028%) is the cheapest paid option. On the maker side the ranking changes: Aster and Lighter Standard are free, and Hyperliquid’s 0.015% maker rate is a third of its taker rate — the reason Hyperliquid scalpers live on post-only and Chase orders. And the big CEXs are not cheaper for a new scalper: Binance, OKX and Bybit charge 0.05%–0.055% taker at their regular tiers. They get cheaper only at VIP volumes (Binance VIP 2 needs $10M of 30-day volume for 0.04%). What CEXs still offer is a fiat on-ramp and account support, in exchange for KYC and custody of your funds. Full schedules: Hyperliquid fees, Lighter fees, Aster fees, DEX vs CEX fees.

5. What a real scalping day costs

Assume 20 round trips a day of $10,000 each: 40 fills and $400,000 of daily notional, or $5.6M over 14 days.

This is where the venues separate. A taker scalper at this pace pays $4,800 to $5,400 a month on Aster or Hyperliquid, $6,000 on Binance and nothing on Lighter Standard. The same scalper working maker orders pays $1,728 on Hyperliquid and nothing on Aster. After about two weeks the $5.6M of rolling volume lifts a Hyperliquid account into tier 1 (0.040% / 0.012%) and an Aster account to the VIP 2 volume bar (0.035% taker, if 500 ASTER is also held).

6. Hyperliquid: fee tiers for high-volume scalpers

Hyperliquid’s tiers use rolling 14-day weighted volume (spot volume counts double), and they matter for scalpers because scalpers climb them:

Source: Hyperliquid docs “Fees” (perpetuals), verified October 19, 2026.

• Referral: code HOLYGRAIL takes 4% off your fees on the first $25M of volume (not on vaults or sub-accounts).
• HYPE staking stacks on top: 5% off above 10 HYPE (Wood), rising to 40% above 500,000 HYPE (Diamond). At tier 0 that moves taker from 0.045% to between 0.0428% and 0.0270%.
• Maker rebates for accounts with more than 0.5%, 1.5% or 3% of 14-day maker volume: -0.001%, -0.002% and -0.003%. Relevant only to professional market makers.

Hyperliquid prices are limited to five significant figures, so BTC trades in whole-dollar ticks and its quoted spread (about 0.12 bp) is wider than Lighter’s or Aster’s $0.10. For a taker that is a fraction of a basis point; for a maker it means each tick is worth a little more. Our maker rebate strategy and order types guide go deeper.

7. Speed and fill quality: which DEX is fastest, and what it costs

None of these venues charges gas per order, so the old “gas per trade” objection to DEX scalping is gone. Speed is the real difference, and the venues publish different kinds of number: Hyperliquid a measured end-to-end latency, Lighter deliberate delays by account type, Aster and dYdX a block time.

Official docs (Hyperliquid hypercore and priority-fee pages, docs.lighter.xyz trading fees, docs.asterdex.com Aster Chain overview, dYdX docs); each figure carries its own check date in our facts registry. Hyperliquid fills that slip are usually capped by order type: TP/SL market orders use up to 10% slippage tolerance and TWAP sub-orders up to 3%, so put a limit price on TP/SL orders in thin markets.

So which is the fastest DEX for scalping? On published numbers, Lighter Premium and Hyperliquid, with Lighter Standard deliberately slower in exchange for zero fees. API limits that bind a scalping bot are in section 8.

What latency costs is adverse selection. A market order that waits 300 ms can arrive after the price has moved; a resting quote that takes 300 ms to cancel can be filled by someone faster just before a move. For a human scalper clicking entries on 1-minute candles, this is small next to reaction time. For a bot quoting both sides, it is the reason to pay for Lighter Premium or to use Hyperliquid.

8. API and high-frequency scalping: rate limits and latency by venue

Once a scalping strategy runs as a bot, the limit that binds first is usually the venue's rate limit or latency, not its fee. A strategy that turns over $1M a day pays $450 at 0.045%, $400 at 0.04% and nothing at 0%. It can also be throttled out of the market by a 60-requests-a-minute cap, or picked off by a 300 ms cancel delay. Here is what each venue publishes, read from the official API docs and live endpoints on September 25, 2026.

Sources: Hyperliquid docs hypercore/overview.md, rate-limits-and-user-limits.md, optimizing-latency.md, priority-fees.md, trading/market-making.md; apidocs.lighter.xyz rate-limits.md (updated October 13, 2026), account-types.md, volume-quota-program.md, get-started.md; docs.lighter.xyz trading-fees.md; Aster fapi exchangeInfo rateLimits and BTCUSDT filters (live, September 25, 2026), docs.asterdex.com overview and the Aster API docs. Lighter's latency figures were re-read on both doc sites on October 6, 2026, when they agreed; its rate limits on the rate-limits page on October 15, 2026.

How Hyperliquid's address budget works in practice. Every USDC you trade adds one action to your allowance, so a bot that places $100 orders needs a 1% fill rate to break even on requests. That makes spam expensive and filled orders cheap. A batch of n orders counts as one request against the IP limit but as n against the address limit. Sub-accounts are treated as separate users, each with its own budget. Cancels keep working after you hit the limit, so you can always flatten. The docs add two more rules. Cancels and ALO (post-only) orders sent at the same moment as IOC/GTC orders almost always execute first. The "fast" cancel flag has no downside except that it cannot cancel trigger orders. Details and code: Hyperliquid rate limits and WebSocket guide.

Lighter sells speed and throughput as a product. Standard's 60 requests a minute is fine for a human or a slow grid bot. It is not enough for quoting. Premium removes the cancel delay and scales order throughput with staked LIT. Its volume quota also rewards trading: $10M of volume adds 5M order transactions. The quota is shared across sub-accounts and never expires. Cancels do not use quota, and one free transaction comes every 15 seconds. The one official colocation statement among these venues is Lighter's AWS Tokyo zone. Tokyo-based bots start close to the sequencer. See the Lighter API guide.

Aster uses Binance-style limits (weight per minute, orders per 10 seconds). New integrations use the V3 API, which requires a completed main-wallet deposit before authenticated endpoints work (error −5050 otherwise, since September 1, 2026). It publishes no latency figure, so measure your own order-to-acknowledgement times before sizing up.

Matching the venue to the strategy, per the published limits:

• Turnover-heavy but latency-tolerant (minute-bar statistical strategies, grids, TWAP-style execution): Lighter Standard if 60 requests a minute is enough, because zero fees can make 5–10 bp edges viable. Otherwise use Hyperliquid, whose allowance grows with every dollar traded.
• Quoting at the touch: Lighter Premium (no cancel delay, 140 ms taker, Tokyo colocation) or Hyperliquid (cancels and ALO prioritised over takers, optional priority fees).
• Two-sided quoting on mid-caps: Aster's 0% maker fee, within 1,200 orders a minute.

Many API traders run on more than one venue and route by edge. None of the three charges gas per order. More in our DEX trading bots guide and Hyperliquid API guide.

9. Which crypto to scalp: spread and depth

The best crypto for scalping is the one where crossing the spread costs least and a normal move is larger than your fees: BTC first, then ETH, SOL and, on Hyperliquid, HYPE. To show the gap, we read Hyperliquid’s impact prices, the price to buy and to sell the funding impact size (20,000 USDC for BTC and ETH, 6,000 USDC for other coins), for the ten most traded perps twenty times between 22:27 and 22:32 UTC on September 26, 2026, a Saturday night, and took the median. The range shows how far each spread jumped in those five minutes:

Hyperliquid info API metaAndAssetCtxs (impactPxs, dayNtlVlm) and meta (maxLeverage), 20 reads 15 seconds apart, September 26, 2026, 22:27–22:32 UTC. Weekend volumes run well below weekdays.

BTC and ETH cost less than half a basis point to cross; HYPE, XRP, SOL and ZEC under one. WLD, ENA and NEAR cost about 4 bp and PUMP almost 7 bp, which is as much as a Hyperliquid taker fee on its own, with a 10x leverage cap and sharper wicks. The ranges matter as much as the medians: NEAR’s spread swung between 0.3 and 9.6 bp within five minutes, while BTC’s barely moved. For larger clips the cross-venue depth snapshot below shows the same order. We pulled all three public order books twice on September 25, 2026 (09:07 and 09:13 UTC) and simulated market buys. “vs mid” is the average fill price above the mid price, including the half-spread; 1 bp = 0.01%.

Hyperliquid l2Book (20 price levels), Lighter orderBookOrders (250 resting orders per side), Aster depth (1,000 levels) on September 25, 2026; ranges show the two snapshots. Volumes 09:09 UTC. Books change by the second: treat this as an order of magnitude.

For a $10,000–$100,000 scalp on BTC or ETH, every venue filled within about one basis point of mid — a quarter or less of a single taker fee. Depth starts to separate the venues at $1M on SOL, where Lighter’s book cost about twice as much as Hyperliquid’s or Aster’s, and on long-tail alts, where Hyperliquid carries far more volume than the other two. See the slippage guide for how to check a book before sizing up.

10. Best time to scalp crypto

Crypto trades around the clock, but scalping works best when the market moves and the books are full. We pulled 30 days of hourly BTC perp candles from Hyperliquid (August 27 to September 26, 2026):

• Busiest: weekdays 13:00–17:00 UTC (9 a.m. to 1 p.m. in New York, 22:00–02:00 in Tokyo), the London–New York overlap. The 14:00 UTC hour averaged $285M of BTC volume.
• Quietest: the Asian afternoon, 03:00–07:00 UTC. The 04:00 UTC hour averaged $54M, about a fifth of the peak.
• Range, which is what a scalper sells: the average hourly high-to-low range was 0.95% at 14:00 UTC and 0.35% at 04:00 UTC. A 0.09% taker round trip eats about a tenth of the peak-hour range but a quarter of the quiet-hour range.
• Weekends: hourly BTC volume averaged $47M against $138M on weekdays, so spreads widen and moves are thinner.

Fees do not change by the hour; spread, slippage and the room for a target after fees do. Scheduled US data (CPI, FOMC) moves the market in seconds and can blow through stops, so either trade those minutes on purpose or stand aside. Hourly funding on Hyperliquid and Lighter settles on the hour. The full hour-by-hour and day-of-week data, plus event times, is in our guide to the best time to trade crypto.

11. Order types for scalpers

From the Hyperliquid order-types docs and app, the Lighter app (v6.0.46) and docs, and the Aster trading page, docs and API, September 25, 2026.

The key tool for a fee-paying venue is the auto-repricing post-only order: Hyperliquid’s Chase and Aster’s Chase follow the best bid or ask and only ever fill as a maker, which turns a 0.045% taker entry into a 0.015% maker entry (Hyperliquid) or a free one (Aster). The second tool is reduce-only IOC for exits that cannot accidentally flip your position. On Lighter Standard, where market orders are free, the logic is simpler: take liquidity and keep your limit orders for take-profits.

12. Funding on short holds

Funding is exchanged at fixed times, so a scalp that opens and closes between two funding times pays none:

• Hyperliquid: every 1 hour, at one eighth of the 8-hour rate; the interest baseline is 0.01% per 8 hours (0.00125% per hour), and the cap is 4% per hour.
• Lighter: every 1 hour, capped at 4% per 8 hours (0.5% per hour).
• Aster: every 8 hours on BTC, ETH and SOL (00:00, 08:00, 16:00 UTC; many alts 1–4 hours). At 09:13 UTC on September 25, 2026 the last BTCUSDT rate was 0.0046% per interval.

Example: a $10,000 BTC scalp that crosses one Hyperliquid funding time at the baseline pays about $0.13, and one crossing Aster’s 16:00 UTC timestamp at 0.0046% pays about $0.46. Both are small next to the $8.64–$9 Hyperliquid taker round trip. Funding only matters when you scalp crowded markets where the hourly rate is many times the baseline; check the market header before you hold through the hour. More in funding rates explained.

13. Leverage and liquidation distance

Maximum leverage on BTC: Hyperliquid 40x (up to $150M), Lighter 50x at any size, Aster 150x up to $300,000 of notional (200x only on the first $400). A rough rule for how far price can move before liquidation is initial margin minus maintenance margin:

• Hyperliquid BTC at 40x: 2.5% − 1.25% = about 1.25%.
• Lighter BTC at 50x: 2% − 1.2% = about 0.8%.
• Aster BTC at 150x: 0.67% − 0.33% = about 0.34%.
• Any venue at 5x: about 18–20% — far outside a normal scalping stop.

Scalpers profit from many small wins, not from one leveraged bet; a wick that stops a 5x trade merely costs the stop, while the same wick liquidates a 150x one and adds a liquidation fee. Most scalpers stay between 3x and 10x. Venue-by-venue tiers are in our DEX leverage tiers comparison.

14. A zero-fee scalping playbook on Lighter

Zero taker fees change how you scalp: market entries and exits cost nothing, so tighter targets become viable and scaling in costs no extra fees. A simple, testable setup:

• Account: Lighter Standard, USDC collateral (sign-up guide, deposit guide). BTC and ETH have the deepest books and 50x caps; SOL (25x) is the most liquid alt. The minimum order is $10.
• Size and leverage: 3–5x; never risk more than 1% of the account per trade (a $50 maximum loss on $5,000).
• Timing: trade the 13:00–17:00 UTC weekday window, when the hourly range is widest; skip the quiet Asian afternoon.
• Entry: a momentum candle on 15-second to 1-minute charts with rising volume, and a spread inside the maximum you set for the pair. Enter with a market order; on Standard it costs nothing.
• Exit: take-profit at +0.3–0.5%, stop at about half the target on the other side. Place both as TP/SL on the position so they trigger on the mark price.
• Discipline: at most three positions at once; stop for the day after a 2% drawdown; keep a journal of win rate and average hold time.

Fee example: $5,000 at 5x controls $25,000 per trade. Fifty round trips at that size cost $0 on Lighter Standard and $1,125 at Hyperliquid’s 0.045% taker rate (50 × 2 × $25,000 × 0.045%). This is a fee comparison, not a return forecast. Entry setups and the break-even table are in the next section.

15. Scalping strategies that survive fees

For a scalp with target T, stop S and round-trip cost c, the win rate you need to break even is (S + c) ÷ (T + S). Applied to three common 2:1 shapes at each venue’s base fees:

Break-even win rate; base tier, no staking discount, slippage and funding excluded. “Code HOLYGRAIL” = the 4% referral discount on your first $25M. Run your own stop and target in the risk reward calculator.

The fill type is part of the strategy: on Hyperliquid and Aster, enter with post-only limits at your level and use market orders only for stops; on Lighter Standard taker entries cost nothing. Three setups that keep the target well above the fee:

• Order-book imbalance at a level. Mark the level (day high or low, round number, range edge) on a 5- or 15-minute chart. When price trades into it, sellers print and the bids at the level keep refilling for several seconds, bid one tick above with a post-only order; stop just beyond the level (typically 0.08–0.15% on BTC in active hours), target the range mid or the next ask cluster and at least 2× the stop. A single book snapshot means little: on Hyperliquid BTC on September 25, 2026 the book within 2 bp of mid flipped from 70% bids to 47% ask-heavy in 16 seconds. What counts is persistence.
• VWAP bands. Anchor VWAP at 00:00 UTC (or at a CPI release) and keep it fixed. On a trend day, buy the first pullback to VWAP or the first lower band with a stop under the pullback low; on a range day with flat VWAP, fade the second band back to VWAP. Skip the fade when the band touch comes with a volume spike and a strong close outside it: that is a breakout.
• Breakout retest. After a 30–60 minute range on the 5-minute chart, wait for a 1-minute close beyond the edge on clearly higher volume, then a return to the broken edge. Place a post-only order there, stop back inside the range by more than a typical 1-minute range (about 0.05–0.1% on BTC), take part at 2× risk and trail the rest. No retest, no trade.

Journal every trade with its setup tag, maker or taker on each side, the fee from the Trade History CSV and the net R multiple: within a few weeks it shows which setup pays after fees. Hourly windows for these setups are in section 10; order types in section 11.

16. edgeX, dYdX, Paradex and Drift for scalping

Other lists name more venues as scalping picks. Here is what their official docs say, and why they rank below the top three for most scalpers:

• edgeX is often labelled “best for scalpers”. Its docs claim “microsecond latency” for its matching engine (edgeX V2 docs; no end-to-end figure published). The fee schedule is the catch: 0.045% taker and 0.04% maker at the regular tier, so a post-only round trip still costs $8 on $10,000. Its top VIP tier drops to 0.02% / 0.015%. BTC leverage goes to 100x. Code HOLYGRAIL works there too, with no published referee discount.
• dYdX: 0.05% taker and 0.01% maker at tier 1, maker rebates at its top tiers, and an order-book chain where, per its docs, order-to-trade takes at least one block (about 800 ms). A solid venue for patient maker scalping, slower for takers.
• Paradex: Retail accounts pay 0% with a 300 ms speed bump on orders and cancels and a cap of 1,000 orders a day. Exceed it and the account is charged like Pro (0.045% taker). It is a zero-fee alternative to Lighter Standard for a low-frequency manual scalper, with a thinner book; where that thin book starts to cost more than Hyperliquid's fee is measured in Paradex vs Hyperliquid.
• Drift is not an option now. It was exploited on April 1, 2026; Drift’s recovery plan puts user losses at $295M, and the Drift program is paused. Details in our Solana perp DEX guide.

17. Verdict by venue

• Hyperliquid — best all-round scalping DEX. The deepest books, the widest market list (178 perps as of October 20, 2026), cancels first in every block, Chase orders for 0.015% maker fills, and fee tiers that scalpers actually reach. Use code HOLYGRAIL for 4% off. Head-to-heads: Hyperliquid vs Lighter, Hyperliquid vs Aster.
• Lighter — best zero-fee scalping DEX. 0% on both sides by default; the right choice for manual BTC/ETH scalping and any strategy dominated by market orders. Upgrade to Premium only if you measure that 300 ms is costing you more than 0.028%.
• Aster — best for maker-only scalping and high leverage on small size. 0% maker at any size, a native Chase order, trailing stops and hidden orders; 0.04% taker. Full comparison with Lighter: Lighter vs Aster.
• edgeX and dYdX are usable but not scalping-first: edgeX charges 0.04% even to makers, and dYdX 0.05% taker at the base tier.
• Binance, Bybit and OKX are the CEX alternative: similar or higher regular-tier fees, a fiat on-ramp and support, but KYC and custody of your funds.

18. FAQ

Disclosure: HOLYGRAIL (Hyperliquid, edgeX), 718610TD (Lighter) and DTLpLc (Aster) are our referral codes; we earn a share of referred accounts’ fees (on Lighter only from Premium and advanced-order fees). Fees, latency figures and order types come from each exchange’s documentation and app; depth, volume and hourly data from their public APIs (September 25–27, 2026). Not investment advice: leveraged scalping can lose money quickly.