DEX Aggregators Explained: How Jupiter, 0x & CoW Swap Route
On 16 November 2026 a 500 ETH sale returned $892,851 via the best Uniswap pool and $898,620 via Velora Delta, but at 1 ETH the gain fell to cents. This guide shows where routing pays.

Introduction
On 16 November 2026, a 500 ether (ETH) sale returned $892,851 through the best single Uniswap pool and was quoted at $898,620 through Velora Delta. A decentralized exchange (DEX) aggregator searches for that gap across pools and market makers, and intent-based swaps hand the search to competing solvers. Both promise the best swap rate in crypto, but the gain depends on trade size, gas and the drift between quote and settlement. This guide compares live quotes from KyberSwap, Velora, CoW Swap and Jupiter with single pools, ranks aggregators by 30-day volume, and explains how routers, batch auctions and Dutch auctions differ. Each measured figure carries its timestamp, and no quote was executed, so the numbers show what the tools offered at one moment.
Key Takeaways
- Jupiter leads 30-day aggregator volume at $12.68B of $63.2B in the 30 days to 15 November 2026; 1inch ranks eighth, and DefiLlama has revised CfoSwap's one-day $139.95B spike on 1 October out of its history.
- On a 500 ETH sale, KyberSwap, Velora and CoW Swap quotes beat the best single Uniswap pool by $5,548 to $5,769 after gas.
- On a 1 ETH sale, the gain shrank to cents: three aggregator quotes beat a single 0.01% Uniswap pool by under $1, and KyberSwap's fell short of it (KyberSwap and CoW Swap APIs, 2026).
- On Solana, choosing the right venue delivered just over half of Jupiter's gain on a 10,000 SOL (Solana's token) sale, and splitting added 0.40% (Jupiter quote API, 2026).
- Intent venues such as CoW Swap and UniswapX sign a floor price, and solvers or fillers carry gas and failed-fill risk.
What Does a DEX Aggregator Do That a Single Pool Cannot?
A decentralized exchange (DEX) aggregator is a router: it prices one swap across many pools and market makers, then builds the path that returns the most tokens. Gas sets its limit, because every extra pool in the path adds execution cost. The aggregator pays off only when the gain exceeds that cost.
One Quote Across Many Venues
One token pair trades in dozens of places at once: Uniswap pools at several fee tiers, Curve, SushiSwap and professional market makers that quote off-chain. A decentralized finance (DeFi) ↗ aggregator queries them all for a single swap and returns one route. 0x states that its API draws on 150+ liquidity sources across Ethereum Virtual Machine (EVM) chains, Solana, Tron and HyperCore (0x docs, 2026). The routes can be wide. On 16 November 2026, a 500 ether (ETH) quote from KyberSwap spread across nine paths over eight venue labels, while the market route from Velora (formerly ParaSwap) for the same trade split into two branches over Fluid, Metric, Titan and Uniswap V4 pools.
Splitting Cuts Price Impact and Every Hop Costs Gas
Price impact grows with trade size, because a pool's price moves ↗ as the trade consumes its liquidity. Splitting one order across pools lets each pool absorb a smaller slice at a smaller price move. Each additional hop adds gas, the cost of running the swap on-chain. KyberSwap's 500 ETH route carried a 4.71M gas estimate, 13 times the 370,452 gas of the best single Uniswap V3 pool: about $1.21 against $0.09 at 0.142 gwei. The split pays when the extra output beats that gap. An $898,000 sale cleared it by about $5,700, while on a sale of about $1,800 the best routes gained less than $1.
Volume data shows which routers the market already trusts with that trade.
Which DEX Aggregators Moved the Most Volume in the Past 30 Days?
Jupiter processed $12.7 billion of swaps in the 30 days to 15 November 2026, the most of any DEX aggregator, ahead of DFlow at $5.85 billion and 0x at $5.78 billion (DefiLlama, 2026). In early October, the same ranking held only after CfoSwap was excluded.
The Ranking Once CfoSwap Is Removed
CfoSwap, an aggregator on BNB Chain, showed $139.95 billion on DefiLlama for the 30 days to 8 October 2026, nearly twice the $72.6 billion that every other protocol in DefiLlama's aggregator view recorded combined. All of it fell on one day, 1 October 2026, and the other days in that window read about zero. DefiLlama has since rewritten the adapter: it now counts the output each swap actually delivered, read from OrderRecord events, because the router does not check the declared input amount against the tokens it pulls, and the revised history shows $123 for 1 October (DefiLlama, 2026). That note points to unchecked declared amounts rather than trading as the source of the spike, which this article had excluded from the start. In the current window, 1inch ranks eighth at $2.9 billion, and the top three venues hold 38% of the $63.2 billion total.
Reading Volume Without Over-Reading It
Aggregator volume counts swaps routed, not liquidity held, so it cannot be added to DEX volume. DefiLlama's Dexs category totals $213.3 billion for the same 30 days, and a trade an aggregator routes also appears as volume on the pools it touches. Names can overlap too: Jupiter's documentation lists DFlow and OKX as routing engines inside its own meta-aggregator, and this article could not confirm whether DefiLlama separates those flows. Rank also says nothing about price. The guide to the 1inch Network ↗ covers the eighth-ranked protocol on its own, and DEX market share by chain ↗ shows where DEX volume sits by chain.
Data current as of November 2026.
Volume ranks the venues, while a quote shows what one venue returns for a specific swap.
Do Aggregators Beat a Single Pool on the Same Swap?
On a 500 ETH sale, all four aggregator quotes beat the best single Uniswap pool by $5,500 to $5,800 after gas. On a 1 ETH sale, the gap shrank to cents, and one aggregator fell short of the pool. The best swap rate in crypto depends on trade size, because size decides how much a better route is worth.
The Test: Same Swap, Same Minute
The test sold wrapped ether (WETH) for USD Coin (USDC) on Ethereum at 10:00:16 UTC on 16 November 2026, block 26,426,852, in three sizes: 1, 100 and 500 ETH. Quotes came from Uniswap's QuoterV2 contract on four V3 fee pools, the KyberSwap Aggregator API, Velora's market and Delta price endpoints, and CoW Swap's quote endpoint. 0x and 1inch require API keys, so neither was measured. Gas was priced at 0.142 gwei and $1,796 per ETH. Nothing was executed. The yardstick, a mid price of $1,795.94, is the V3 0.05% pool's spot price. It is a reference point, not a fair-value claim: a quote above it means liquidity elsewhere priced better than that one pool's marginal price.
At 500 ETH Every Aggregator Beat the Best Single Pool
The best single pool, the V3 0.05% pool, returned $892,851 net of gas, 0.57% below mid. KyberSwap returned $898,541, which is $5,690 more after a $1.21 gas estimate. Velora's market route returned $5,548 more. KyberSwap's route labels name Uniswap V4 and Fluid pools beside proprietary market-maker venues labelled Kipseli, Tempest and Metric. The two intent-based quotes split: Velora Delta returned $5,769 more, the best result, and CoW Swap $5,638 more, behind KyberSwap; all four quotes sat 0.05% to 0.07% above mid with gas already deducted. CoW's documentation credits private market-maker inventory for improvements of this kind (CoW Protocol docs, 2026). A second snapshot 165 seconds later reversed the order: Velora's market route +$5,009, KyberSwap +$4,904, Delta +$4,728, CoW +$4,641.
At 1 ETH the Gap Shrank to Cents
For a 1 ETH sale, the V3 0.01% pool returned $1,796.57 net of gas. Velora Delta returned $1,797.27, Velora's market route $1,797.20 and CoW $1,796.77, so three aggregators beat the pool by $0.19 to $0.70. KyberSwap returned $1,795.97, $0.60 below the pool: its quote was lower before gas, and its 810,495 gas estimate cost $0.21 against $0.03 for the pool. The second snapshot kept the pattern, with Velora's market route ahead at $1,798.20 and KyberSwap again behind the best pool. At 0.14 gwei, gas barely separates the routes, and the gap between the best and worst aggregator quote stayed under $1.60 in both snapshots. At 100 ETH every aggregator beat the best pool, $179,316, by $382 to $420 in the first snapshot (KyberSwap, Velora and CoW Swap APIs, 2026).
Data current as of November 2026.
Solana shows the same pattern at a different scale, with Jupiter pitting four routing engines against each other.
How Does Jupiter Find the Best Rate on Solana?
Jupiter, the largest DEX aggregator by 30-day volume at $12.7 billion in the 30 days to 15 November 2026 (DefiLlama, 2026), runs four routing engines against each other and keeps the best price. A same-minute check on 16 November 2026 shows where its gain comes from: picking the right venue delivered just over half of it, and splitting across several venues the rest.
Jupiter Runs Four Routing Engines
Jupiter's Swap API v2 acts as a meta-aggregator in which Metis, JupiterZ, DFlow and OKX compete for each order (Jupiter docs, 2026). Metis is Jupiter's on-chain router, with multi-hop and multi-split swaps across Solana venues. Jupiter also reports a self-learning mechanism that sidelines underperforming sources. JupiterZ is a request-for-quote (RFQ) system where market makers quote off-chain. Jupiter says it often beats on-chain routing by 5 to 20 basis points on major pairs, a claim this test could not check because the full API needs a key. JupiterZ market makers keep last-look rights, a final chance to decline a fill, and its transactions cannot be modified once Jupiter returns them.
Measured on Solana: Venue Choice Beats Splitting
Jupiter's keyless Swap V1 quote endpoint, which uses the Metis router alone (Jupiter docs, 2026), priced sales of SOL, Solana's native token, for USDC at 10:00 UTC. At 10,000 SOL, about $708,000, all routes returned 708,140 USDC, split across GoonFi V2, BisonFi, Scorch and JupLend AMM, the venue labels Jupiter returned. Orca's Whirlpool pools alone returned 701,673, so routing gained 0.92%, or 6,468 USDC. The best single venue Jupiter found, a direct route through Scorch, returned 705,299, 0.40% below the four-leg split (Jupiter quote API, 2026). Choosing the venue delivered just over half of the gain, and splitting added the rest. At 10 and 1,000 SOL the direct route quoted slightly more than all routes combined, so splitting added nothing at those sizes.
Data current as of November 2026.
Both measurements are quotes, and a quote is not the amount that lands in the wallet.
Why Can the Executed Price Differ From the Quoted Price?
A quote is the router's simulation of one swap against the pool state at one block. The trade lands later, against a changed state. Slippage tolerance governs the gap: the swap either executes within the limit or reverts.
A Quote Is One Block's Estimate
Quote drift appeared in the test itself. Between the first snapshot at 10:00:16 UTC and the second at 10:03:01 UTC, 14 blocks later, the V3 0.05% pool's mid price rose from $1,795.94 to $1,798.18 (Uniswap V3 pool contract, 2026). KyberSwap's quote for 500 ETH rose $327 over those 165 seconds, and the CoW Swap quote for 1 ETH moved from 4.6 basis points above mid to 1.9 below it. A router quote reserves nothing: the transaction executes against whatever the pools hold when a validator includes it. An intent-based quote works differently, because the signed order carries a floor, and UniswapX cosigners update auction parameters to current prices without crossing the slippage the swapper signed (Uniswap docs, 2026).
Slippage Settings Set the Floor, Not the Price
Slippage tolerance is the minimum output a trader accepts, expressed as a percentage of the quote. 1inch defines it as the smallest token amount the trader will take if the price moves. Set too low, the transaction fails with a "Min return not reached" error and the gas is spent (1inch Help Center, 2026). Set too high, the swap can settle at a wider range of prices. On Jupiter's 1 SOL quote, a 0.5% setting put the floor at 70.54 USDC against a 70.90 USDC quote. On the 500 ETH sale a 0.5% floor would allow a shortfall of about $4,500, most of the $5,690 that KyberSwap's route gained. Intent venues write the floor into the signed order, and UniswapX decays its auction price down to exactly that floor (Uniswap docs, 2026).
A signed floor is the first thing that separates an intent from a routed transaction.
How Do Intent-Based Swaps and Solver Auctions Work?
Intent-based swaps replace the routed transaction with a signed statement of the outcome wanted. Competing solvers, fillers or resolvers then build the execution and carry its cost and risk. CoW Swap, UniswapX, Velora Delta and 1inch Fusion all work this way, with different auction rules.
The User Signs an Outcome, Not a Path
An intent is a signed message stating assets, amounts and a limit price, not an executable transaction (CoW Protocol docs, 2026). UniswapX orders are signed through Permit2, which releases tokens only if the amounts sent and received match the order's decay curve (Uniswap docs, 2026). Third parties compete to execute it, and the settlement contract checks the signature and the limit before any tokens move. Solvers hold no user approvals on CoW Protocol; transfers run through its vault relayer contract. Gas follows from that design. CoW Swap takes network costs from the sell token, and UniswapX fillers pay gas and charge nothing for failed swaps (CoW Protocol and Uniswap docs, 2026). In the 16 November test, CoW's network cost was 0.000065 ETH on the 1 ETH quote and 0.000067 ETH on the 500 ETH quote: about $0.12 either way.
CoW Swap Runs a Batch Auction
CoW Swap gathers orders off-chain into batches and auctions each batch to solvers. In the fair combinatorial auction, solvers bid on single orders or on groups, and the protocol picks the combination that maximizes surplus for the traders (CoW Protocol docs, 2026). Orders for the same pair in one batch settle at one uniform clearing price, so reordering them earns nothing. Opposite orders can match peer to peer, a coincidence of wants that skips pool fees. The protocol also requires settlement at or above the Ethereum best bid and offer, the price on-chain automated market makers (AMMs) offer (CoW Protocol docs, 2026). Solvers are bonded and paid in COW tokens. The protocol charges a 2 basis point volume fee on standard assets and 0.3 basis points on correlated assets such as stablecoins.
UniswapX and Velora Delta Also Run Auctions
UniswapX uses a Dutch auction. On Ethereum, the quoter that wins a request-for-quote round gets about 24 seconds of exclusive fill rights. The price then decays for about a minute until a filler takes the order or it expires (Uniswap docs, 2026). ParaSwap renamed itself Velora on 3 April 2025 and launched Delta v2.5 the same day, where agents, meaning solvers and market makers, compete to execute intents (Velora, Cointelegraph, 2025). 1inch Fusion runs a similar Dutch auction among staked resolvers who cover the gas, as the 1inch guide linked above details.
Data current as of November 2026.
Those mechanics carry costs and failure modes of their own.
What Do Intents Cost, and When Does a Router Still Win?
Intent-based swaps remove gas risk and failed-transaction fees from the trader, but they add a protocol fee, a wait and an expiry. For small trades, a router that fills at once can still return more.
What Intents Cost
CoW Swap's 2 basis point volume fee is built into its quote, so the $898,489 quote for 500 ETH is already net of it. The market-order quote-improvement fee takes 50% of any price gain over the quote, capped at 0.98% of volume (CoW Protocol docs, 2026). Velora's Delta quote subtracts a gas cost from the output: $0.12 on the 100 ETH quote. Time is the other cost. UniswapX's Ethereum auction decays for about a minute before expiry, and a 1inch Fusion order that no resolver fills expires and must be resubmitted at no charge (Uniswap docs; 1inch Help Center, 2026). These costs buy a signed floor and no gas lost on a failed fill, which a failed router swap does not offer.
Where a Router Still Wins
A router settles in one transaction that the trader controls, which matters for custom builds. Jupiter's Router path returns raw instructions for custom transactions, while JupiterZ transactions cannot be altered (Jupiter docs, 2026). Small size matters too. In the 1 ETH test, a single V3 0.01% pool beat KyberSwap's route on net output in both snapshots and came within $0.70 of the best aggregator quote. On Solana the 10 SOL quotes sat within 0.07% across all three methods, so small trades gain little from either approach. A router also fills at once, with no exclusivity window or decay to wait through. For a single-pool swap done by hand, the step-by-step guide to using Uniswap ↗ walks through the steps. Routers and intents answer one question differently: who carries the risk between quote and settlement.
Conclusion
Aggregation is a size-dependent tool, not a free upgrade. A router buys better prices with more hops and more gas, and with gas near 0.14 gwei the quotes show the gain tracking size: about $5,700 on a 500 ETH sale and less than $1 on 1 ETH. Intent-based swaps hand the same search to competing solvers and fillers, who absorb gas and execution risk in exchange for a fee, a wait or an expiry.
The difference between a good outcome and a bad one is invisible at quote time. A router quote reserves nothing, and 165 seconds moved KyberSwap's 500 ETH quote by $327, far more than the whole spread among aggregators on a 1 ETH sale. A loose slippage setting then converts that drift into a real loss: a 0.5% floor on the largest sale allowed a shortfall of about $4,500 (own calculation, 2026). Volume rankings and route counts say little about this risk. The floor does.
Before swapping, compare one aggregator quote with the best single pool at the size being traded, net of gas. Set slippage tolerance as a floor you can afford, and prefer an intent venue for large orders, where a signed minimum and a gas-free failure matter. For a swap of a couple of thousand dollars, a single deep pool came within $1 of the best route in the 1 ETH test. Rerun the quote immediately before signing, because the price on screen is already stale.
Quick Stats
- $12.68B — Jupiter's 30-day aggregator volume to 15 November 2026 (DefiLlama, 2026), the highest listed
- $139.95B — CfoSwap's single-day volume on 1 October 2026 as first reported, since revised away by DefiLlama
- +$5,769 — Velora Delta's quote advantage over the best single pool on 500 ETH
- 13× — KyberSwap's gas estimate against one Uniswap V3 pool on that sale
- 0.92% — Jupiter's all-routes gain over Orca Whirlpool alone on 10,000 SOL
- 2 bps — CoW Protocol's volume fee on standard assets
Data current as of November 2026.
FAQ
?Does a DEX aggregator charge its own fee on top of the pool fee?
It depends on the product. KyberSwap's 1 ETH route response carried no extra fee on 16 November 2026. Jupiter's Meta-Aggregator path applies a Jupiter platform fee, and integrators can add referral fees on top (Jupiter docs, 2026). CoW Protocol charges a 2 basis point volume fee on standard assets, built into its quote. Compare quotes net of every fee, since the quoted output already includes them.
?Is a DEX aggregator the same thing as a DEX?
No. A DEX holds liquidity in its own pools ↗ or order books, while an aggregator routes a swap into DEXs and market makers. Trades routed through an aggregator also appear as volume on the pools they touch, so DefiLlama lists the two separately. Some products blur the line: CoW Swap and UniswapX run their own auction protocols on top of the same liquidity.
?How is CoW Swap different from 1inch Fusion?
Both let third parties execute a signed order and cover the gas. CoW Swap collects orders into batches, runs a solver auction per batch, and settles same-pair orders at one uniform clearing price. 1inch Fusion auctions each order separately as a Dutch auction, where the rate falls until a staked resolver fills it (CoW Protocol docs; 1inch Help Center, 2026).
?What happens to an intent-based swap that no solver or filler picks up?
The order expires without a trade. UniswapX advertises no cost for failed swaps, and a 1inch Fusion order that no resolver fills can be resubmitted at no cost (Uniswap docs; 1inch Help Center, 2026). On Ethereum the UniswapX auction decays for about a minute before expiry. The trader loses time and possibly a better price, but pays no gas for a reverted transaction.
?Which aggregator quote APIs can a developer call without an API key?
On 16 November 2026 the KyberSwap Aggregator, Velora, CoW Swap quote and Jupiter Swap V1 endpoints answered without a key. 0x returned "No API key found in request," 1inch returned "Unauthorized" and Odos returned a Cloudflare error. Access rules change, and Jupiter's current Swap API v2 requires a key (Jupiter docs, 2026), so check each provider's documentation.
?Is a swap through an aggregator riskier than swapping on one DEX?
An aggregator adds its own router contract to the path, so tokens and approvals touch one more contract than in a direct swap. In exchange, routes can include more pools and market makers. Smart-contract risk ↗ and approval hygiene therefore apply to the router as well as the pools, and a reverted swap costs the same gas as any failed transaction.
?Do aggregators handle tokens that charge a transfer tax?
Support varies by provider. 0x documents real-time detection and calculation of buy and sell taxes (0x docs, 2026). The other aggregators were not tested here, so a trader should check the provider's documentation and the minimum output on the confirmation screen, because a tax lowers what arrives even when the quote looks correct.
References / Sources
Platform & Company Data
On-chain reads, public quote endpoints and volume data pulled on 16 November 2026.
- DefiLlama: Aggregator volume and CfoSwap daily data (defillama.com, Nov 2026)
- KyberSwap: Aggregator API route quotes (aggregator-api.kyberswap.com, Nov 2026)
- Velora: Prices and Delta API quotes (api.paraswap.io, Nov 2026)
- Jupiter: Swap V1 quote endpoint (lite-api.jup.ag, Nov 2026)
- Ethereum RPC: Uniswap V3 QuoterV2 and pool reads (ethereum-rpc.publicnode.com, Nov 2026)
- Cointelegraph: ParaSwap rebrands to Velora (cointelegraph.com, Apr 2025)
Academic & Technical
Protocol documentation for routing, auction and fee mechanics.
- CoW Protocol: Auctions, solvers, intents, fees and quote API (docs.cow.fi, 2026)
- Uniswap: UniswapX overview and auction types (developers.uniswap.org, 2026)
- Jupiter: Swap API v2 overview and routing (dev.jup.ag, 2026)
- 0x: Docs introduction (0x.org, 2026)
- 1inch: Fusion and slippage help articles (help.1inch.com, 2026)
- Velora: Unveiling Velora, the next era of ParaSwap (medium.com, Apr 2025)
Original CoinPaprika guide.