Ranked & Reviewed
Best Decentralised Exchanges
Non-custodial exchanges ranked on execution quality, contract security and real liquidity rather than posted swap fees.
No money changes hands. This is a non-commercial project: no advertising, no sponsored placements, and no affiliate or referral links. Links to the services below earn us nothing. Every overall score is the plain average of the criteria shown on the page, so you can check the arithmetic yourself — see our rating methodology and how the site is funded.
Uniswap publishes its fee tiers plainly — 0.30% flat on v2, 0.05%/0.30%/1% on v3, freely settable on v4 — and states that core contracts cannot be upgraded, paused or reversed. Its licence is weaker than its documentation claims.
Pros
- v3 tiers published as 0.05%, 0.30% and 1%; v2 a flat 0.30%
- Core contracts stated to be non-upgradeable, unpausable and non-reversible
- Deployed across 24 networks per official deployment docs
- Audits published as PDFs from OpenZeppelin and Trail of Bits
Cons
- v4-core ships under Business Source License 1.1, which is not an open-source licence
- User documentation describes the protocol as open-source and GPL-licensed, contradicting the licence files
- v4 lets pool creators set any fee from 0% to 100%, so the tier guarantee does not carry over
Balancer is released under GPL-3.0 — a recognised open-source licence, unlike most of this category — and states that its core contracts are immutable by design, with pool updates requiring users to migrate voluntarily.
Pros
- GPL-3.0 in both v2 and v3 monorepos, a recognised open-source licence
- Core contracts described as immutable by design; updates require voluntary user migration
- Fee bounds published: 0.001%-10% for weighted pools, 0.0001%-10% for stable pools
- Deployed across 13 named networks
Cons
- Publishes bounds rather than discrete fee tiers, so the actual rate is per-pool
- Only one audit firm, Certora, is named on the security page
- v3 imposes no swap fee limit at the vault level
PancakeSwap publishes the longest audit list in this comparison — ten named firms — and licenses its Infinity core under GPL-2.0. Its fee documentation gives caps rather than the discrete tiers a trader can look up.
Pros
- Ten auditing firms named including Cyfrin, BlockSec, PeckShield, SlowMist, OtterSec, Zellic and Halborn
- Infinity core released under GNU General Public License v2.0
- Static pool fees fixed at creation and cannot be changed afterwards
- Cross-chain swaps across seven named networks
Cons
- Fee documentation publishes caps (up to 100% CLAMM, 10% LBAMM) rather than usable tiers
- The v3 contracts repository has no LICENSE file at its root
- Contract upgradeability is not stated
SushiSwap publishes its v2 fee and its split — 0.3% per trade, 0.25% to liquidity providers and 0.05% to SUSHI holders — plus v3 tiers, and claims 38 supported chains. Its licence position and audit disclosure are both muddled.
Pros
- v2 fee published with its split: 0.25% to LPs, 0.05% to SUSHI holders
- v3 tiers of 0.01%, 0.05%, 0.3% and 1%
- 38 chains stated, the widest deployment in this comparison
Cons
- The v3-core licence file names Uniswap Labs as licensor and Uniswap V3 Core as the licensed work
- v2-core and the main monorepo have no LICENSE file on the branches checked
- No audit page or named auditing firm was found
- The chain list is self-dated July 2025, a year stale at the time of checking
Raydium publishes four exact concentrated-liquidity fee tiers — 0.01%, 0.05%, 0.25% and 1.00% — and releases its contracts under Apache 2.0. It is Solana-only, and neither audits nor upgradeability are documented.
Pros
- Four exact CLMM fee tiers published: 0.01%, 0.05%, 0.25%, 1.00%
- Apache License 2.0 on both the CLMM and CP-swap repositories
- Tiers given with their raw AmmConfig values, so they can be checked on chain
Cons
- Solana only — no multi-chain deployment
- No audits are named on the documentation or repository READMEs
- Contract upgradeability is not stated
Aerodrome states that the protocol is immutable but that its factories remain upgradable — an unusually candid distinction. Fees are given only as a 3% ceiling set per pool by a fee manager, and the licence is BUSL-1.1.
Pros
- States plainly that the protocol is immutable while factories are upgradable
- Fee ceiling published: a maximum of 3% per pool
- MixBytes audit named with dates, September to October 2025
Cons
- No fee tiers published, only a maximum and a note that fees adjust with volatility
- Business Source License 1.1, not an open-source licence
- Deployed on Base and Optimism only
- The official docs site refused automated access, so facts come from the GitHub organisation
Curve is deployed across 24 chains and names four auditing firms, but its repositories carry a licence stating that no right of reproduction or distribution is granted, and its fee pages could not be reached.
Pros
- Deployed across 24 named chains, joint widest coverage here
- Four auditing firms named: ChainSecurity, MixBytes, Quantstamp and StateMind
- Admin fee documented as a constant 50% of the trade fee
Cons
- The LICENSE file states code is published for informational purposes only, with no licence or right of reproduction or distribution granted or implied
- GitHub classifies the repositories as NOASSERTION rather than any recognised licence
- The fee pages returned 403 or rendered empty, so swap fee tiers could not be verified
- Contract upgradeability is not stated
Orca publishes nine fee tiers from 0.01% to 2.00% along with the exact split — 87% to liquidity providers, 12% to the DAO treasury, 1% to a climate fund. In February 2025 it moved off Apache 2.0 to a revocable, non-commercial licence.
Pros
- Nine fee tiers published, from 0.01% to 2.00%
- Fee split published precisely: 87% LPs, 12% DAO treasury, 1% Climate Fund
- Tier parameters documented alongside tick spacing
Cons
- Relicensed on 27 February 2025 from Apache 2.0 to the bespoke Orca License
- The new licence is revocable and limited to authorised non-commercial uses
- It explicitly excludes any use benefiting a competitor, including other DEXs and AMMs
- No audits named; upgradeability not stated
1inch releases its limit order protocol under the MIT licence, one of the cleanest positions in this category. It does not publish its fees as percentages, describing instead a swap surplus retained by the API provider and infrastructure fees covered by the taker.
Pros
- MIT License on the limit order protocol and the legacy protocol repository
- MIT is a recognised open-source licence with no change date or revocation
Cons
- No fee percentage is published; the help page describes swap surplus and infrastructure fees without figures
- Supported chains could not be verified — the business documentation portal rendered empty
- No audits named on the pages we read
- The former fees article now redirects to an unrelated page
Jupiter states plainly that there are zero protocol fees on its swap by default, with integrators able to add their own platform fee. The on-chain aggregator program itself is not published, so the code cannot be inspected.
Pros
- States zero protocol fees on Jupiter Swap by default
- Integrator fees documented with a worked example, 20 bps for 0.2%
- The default cost position is the clearest of any aggregator here
Cons
- No repository containing the core on-chain aggregator program is published
- Only SDKs and clients are public, most without a licence
- A formal supported-chains statement was not found
- No audits named on the pages we read
At a glance
| Service | Score | Best for |
|---|---|---|
| Uniswap | 8.8 | Published tiers and genuinely immutable core |
| Balancer | 8.4 | GPL-3.0 licensed and immutable by design |
| PancakeSwap | 7.6 | Ten named auditors and a GPL licence |
| SushiSwap | 6.2 | Widest chain deployment, thinnest current disclosure |
| Raydium | 6.0 | Exact fee tiers under a clean Apache licence |
| Aerodrome | 5.4 | Honest about which parts stay upgradeable |
| Curve | 5.4 | Four named auditors, but the code grants no rights |
| Orca | 4.6 | Detailed fee tiers under a revocable licence |
| 1inch | 4.4 | MIT-licensed aggregator with unpublished fees |
| Jupiter | 4.4 | Zero protocol fees, unpublished core code |
A decentralised exchange lets you trade without handing custody to anyone, settling each swap on chain against a pool or an order book. That removes counterparty risk and replaces it with contract risk, so audits, upgrade keys and oracle design carry real weight here.
Posted fee tiers tell you little on their own. What a trade costs in practice is the fee plus price impact plus whatever the router gives away in slippage, so we look at realised execution on typical sizes instead of the advertised rate.
How we score this category
A decentralised exchange asks you to trust code rather than a company, so this rubric scores what can be read: the fee the protocol publishes, the licence the code actually carries, whether the contracts can be changed after deployment, and whether audits are named. Licence is treated strictly — a public repository is not an open-source licence, and in this category the gap between the two turns out to be wide.
Every service in this ranking is scored 0–10 against the same 5 criteria, each carrying equal weight (20% apiece). The overall score is their plain average, rounded to one decimal place, and each review prints the individual numbers and the arithmetic so you can reproduce the result yourself.
What each criterion covers
- Fee transparency20% of the score
- Whether swap fees are published as figures a user can look up before trading, including how the fee is split. A protocol that publishes discrete tiers scores above one that publishes only an upper bound, and well above one whose fee page cannot be reached at all.
- Licence & code openness20% of the score
- The licence in the repository's LICENSE file, read directly. A recognised open-source licence scores highest; a delayed-open licence such as BUSL scores mid-table; a revocable or rights-reserving licence scores low even where the code is visible.
- Immutability & governance20% of the score
- Whether the core contracts can be upgraded, paused or altered after deployment, and who holds that power. Protocols that state their position plainly — including admitting which parts remain upgradeable — score above those that stay silent.
- Chain coverage20% of the score
- The networks the protocol states it is deployed on, taken from its own deployment documentation rather than an aggregator.
- Audit disclosure20% of the score
- Whether audit reports are published and the auditing firms named. An unnamed audit cannot be checked, and no audit page at all is scored as the absence it is.
What does not affect a score
Nothing commercial, because there is nothing commercial to affect it. This is a non-commercial project: no advertising, no sponsored placements, and no affiliate or referral links anywhere on the site. Links to the services above earn us nothing, so no provider can buy a score, a rank, or an entry in this list. Where two services finish level, they are listed alphabetically rather than ordered silently.
The process across all categories, including how often we re-check, is set out in our ratings methodology.
Rankings are editorial. Nothing here is financial advice. Editorial policy · How we score · How we're funded.