Ranked & Reviewed · 9 services
Best Crypto Bridges
Nine cross-chain bridges, ranked on five published criteria: fee transparency, trust model disclosure, licence, chain coverage and audits. The marks come from desk research checked in November 2026.
At a glance
| # | Service | Score | Best for |
|---|---|---|---|
| 1 | Axelar | 8.6 | Open-source code and the deepest dated audit archive among bridges |
| 2 | Wormhole | 8.6 | A numbered guardian quorum and a deep, named audit record |
| 3 | Celer cBridge | 8.4 | Open-source contracts with a stated fee range and threshold |
| 4 | deBridge | 7.8 | Per-chain flat fees and a dated audit archive |
| 5 | Chainflip | 7.6 | Native-asset swaps with a published custody threshold and fee table |
| 6 | Across | 7.2 | The clearest statement of a bridge's security assumption |
| 7 | Stargate / LayerZero | 6.6 | Broad chain list and named audits at every protocol stage |
| 8 | Squid | 6.0 | Named, linked audits and wide chain coverage for routed swaps |
| 9 | Synapse | 5.8 | MIT-licensed code and a documented optimistic relay network |
Best Crypto Bridges, ranked by score
Axelar scores 8.6/10 in bridges, unchanged and level with Wormhole at the top of the category: Apache 2.0 code, 61 dated audit reports from ten firms and a clear stake-weighted validator model. It publishes fee components but no figures. Based on desk research of its docs and repositories on 6 November 2026.
Pros
- Audit disclosure 10: 61 dated reports from ten named firms in a public repository
- Licence & code openness 10: axelar-core under the Apache License 2.0
- Trust model disclosure 9: validator votes and stake-weighted key shares explained
- Chain coverage 9: 31 non-Cosmos chains in the mainnet config, 80+ claimed
Cons
- Fee transparency 5: base, execution, L1 data and express fees described with no amounts
- Per-chain-pair base fee schedule is not published; cost is known only from a quote
- Signing threshold is not stated as plainly as the category's clearest trust-model disclosures
Wormhole scores 8.6 out of 10 in bridges, level with Axelar at the top: its 13-of-19 guardian model is documented with what each quorum can do, ten named firms performed its 29 audits, and the code is Apache 2.0. Fees are the gap: no transfer fee is published. Desk research, checked November 6, 2026.
Pros
- Trust model 10: 13 of 19 guardians to pass messages or upgrades; 7 can censor; no emergency pause
- Audit disclosure 10: 29 audits by ten named firms, with reports public
- Licence 9: Apache License 2.0 on the core repository
- Chain coverage 9: per-product mainnet lists, 27 chains for WTT and 33 for NTT
- Documents its own February 2022 Solana exploit and the Governor delay built afterwards
Cons
- Fee transparency 5: no transfer fee published for WTT, NTT or Portal
- Solana message fee and relay delivery costs are not priced; delivery is quoted per request
- No consolidated chain count; the docs say listed chains may not be currently connected
- Licence is a short-form Apache notice that GitHub does not recognize
Celer cBridge scores 8.4/10 in bridges, up from 6.2: it links five audits from three firms, states the two-thirds stake threshold behind its security and gives a 40+ chain count. Its docs give two different protocol fee ceilings. Based on desk research on 6 November 2026.
Pros
- Licence & code openness 10: sgn-v2-contracts under GPL-3.0
- Trust model disclosure 9: funds at risk only if over two-thirds of staked value colludes
- Chain coverage 9: 40+ chains stated; 68 entries in the public gateway API
- Audit disclosure 7: five linked reports from CertiK, PeckShield and SlowMist
Cons
- Fee transparency 7: FAQ gives a 0%–2% protocol fee, the fee section 0%–0.5%
- No audit is dated, and the validator-node audit is listed as in progress
- Quarantine thresholds, delay length and guardian operators are not stated
- No legal entity or team is named on Celer's pages
deBridge scores 7.8/10 in bridges, up from 5.4: it publishes a flat fee for each of 18 chains plus 4 bps, and a dated archive of 33 audit reports. Its docs do not state the validator count or signature threshold behind transfers. Based on desk research on 6 November 2026.
Pros
- Fee transparency 9: flat fee per chain for 18 chains, plus 4 bps, with a worked example
- Audit disclosure 9: 33 dated reports from Halborn, Zokyo, Ackee and Neodyme in a public repo
- Licence & code openness 8: BUSL change date has passed; change licence is GPL-2.0-or-later
- Warns integrators that fees can change and must be queried from the contract
Cons
- Trust model disclosure 6: validator count and signature threshold are not stated
- Chain coverage 7: 18 chains in the fee table, fewer than several peers
- The LICENSE file still shows BUSL text, though its change date has passed
- Newest audit report in the archive is dated December 2024
Chainflip scores 7.6/10 in bridges, up from 4.6: its docs now state a 100-of-150 validator signature over every vault and itemize each fee, and its code is Apache 2.0. Eight chains and no audit index of its own hold it back. Based on desk research of its docs and the Trail of Bits library on 6 November 2026.
Pros
- Trust model disclosure 10: 100-of-150 FROST threshold and governance override keys documented
- Fee transparency 9: 0.10% network fee, 0.10%–0.15% pool fees and Boost range published
- Licence & code openness 10: chainflip-backend under the Apache License 2.0
- Swaps settle in native assets, including BTC, with no wrapped token on arrival
Cons
- Chain coverage 5: eight mainnet chains enumerated
- Audit disclosure 4: no audit page; docs cite audits without naming firms or linking reports
- Deposit gas and destination broadcast fees vary by chain with no published table
- A 3-of-6 governance key can put the network into Safe Mode and pause core functions
Across scores 7.2/10 in bridges: it states its optimistic security assumption in plain terms and links eight OpenZeppelin audit reports. Its fees are a formula with no published rates, and the code is under BUSL until 2027. Based on desk research of its docs on 6 November 2026.
Pros
- Trust model disclosure 10: bonded bundles, UMA oracle disputes, one honest disputer suffices
- Chain coverage 9: 26+ chains stated, including Ethereum, major L2s, Solana and Tron
- Audit disclosure 8: eight linked OpenZeppelin reports, covering Across V2, V3 and UMA
- Fee components explained, including when the LP fee falls to zero
Cons
- Fee transparency 5: no fixed rate; LP curve parameters vary by route and are not listed
- Licence & code openness 4: BUSL-1.1 until 27 July 2027, not an open-source licence today
- Audit page shows no dates and names the firm only through its report links
Stargate scores 6.6 out of 10 in bridges, seventh of nine: 46 networks enumerated in its V2 docs, named audits at every stage and a candid DVN security model. V2 fees are dynamic with no published rate, and neither Stargate's BUSL nor LayerZero's LZBL licence is open source. Desk research, checked November 6, 2026.
Pros
- Chain coverage 9: 46 distinct networks listed with token addresses in the V2 technical reference
- Audit disclosure 9: Zellic, Quantstamp, Zokyo (V1) and Zellic, OtterSec (V2) reports linked
- Trust model 8: DVN threshold model, with Stargate's and Nethermind's DVNs named
- LayerZero states its weak points: default single operator, single-DVN chains
Cons
- Fee transparency 5: V2 fees are dynamic per pathway, with no rate published
- Fee allocation still assigns 1/6 to veSTG holders, a role removed by the 2025 buyout
- Licence 2: Stargate V2 BUSL names no Change License; LayerZero's LZBL-1.2 says it is not open source
- No per-route DVN set or threshold published for Stargate
Squid scores 6.0 out of 10 in bridges: nine audits by four named firms with linked reports, and more than 100 chains claimed. Its router code carries no published licence, its frontend convenience fee has no published rate, and its TEE settlement model names no weak points. Desk research, checked November 6, 2026.
Pros
- Audit disclosure 8: nine audits by Ackee, Consensys Diligence, Kaden and n-Var, with dates, scope and reports
- Chain coverage 9: 100+ chains claimed, with a 39-entry supported-chains table by bridge type
- No protocol fee and no fee on CCTP legs, both stated in the docs
- Squid Intents settlement flow documented: deposit, RFQ auction, fill, TEE-verified release, 15-minute refund
Cons
- Licence 2: no public router-contract repository; SDK and widget repositories carry no licence
- Fee transparency 6: frontend convenience fee on some routes, with no rate or route list published
- Trust model 5: no TEE operator, failure case or per-route trust model stated
- Newest audit is from September 2024, and none covers the off-chain TEE system
Synapse scores 5.8 out of 10 in bridges, ninth of nine: MIT-licensed code, a 19-network chain list and a documented Intent Network that admits Synapse is its sole Guard. It publishes no fee figures, and its audit reports sit unlinked in a code repository. Desk research, checked November 6, 2026.
Pros
- Licence 9: MIT LICENSE on the sanguine monorepo; MIT SPDX headers on the bridge contracts
- Trust model 6: Intent Network escrow, proof, dispute period and Guard checks documented
- Weak point stated: Synapse is currently the sole Guard and sole Canceler
- Chain coverage 7: 19 networks listed with their tokens on the Chains & Tokens page
Cons
- Fee transparency 2: no rate, minimum or split published; the fee is only inside each quote
- Audit disclosure 5: reports by Trail of Bits, Zellic, ChainLight and Aleph_v are not linked from the docs
- Pool-route security is not explained beyond assets being moved to the destination chain
- No dispute-period or cancel-delay durations are published
Best crypto bridges by use case
- You want to know the fee before you ask for a quote: deBridge scores 9 for fee transparency. It publishes a flat fee for each of 18 source chains, paid in that chain's gas token (0.001 ETH on Ethereum, 0.015 SOL on Solana), plus 4 basis points of the amount sent. Chainflip also scores 9: its docs itemize a 0.10% network fee (0.01% on stablecoin swaps, $0.50 minimum) and pool fees of 0.10% to 0.15% per leg.
- You want a stated signing threshold: Wormhole scores 10 for trust model disclosure. Its 19 guardians need 13 signatures for a valid message, and the docs say 7 can censor. Chainflip also scores 10: 100 of its 150 validators must sign for any vault, and the docs say a 3-of-6 governance key can trigger Safe Mode.
- You prefer an optimistic design over a signer set: Across scores 10 on the same criterion. Transfers settle in bonded bundles on Ethereum, which can be disputed through UMA's Optimistic Oracle. It drops to 7.2 overall on a fee formula with no published rates and a BUSL licence that runs until 27 July 2027.
- You want open-source code and a deep audit record: Axelar scores 10 on both. The code is under Apache 2.0, and the audit archive holds 61 dated reports from ten firms, November 2021 to August 2026. The GPL-3.0 code of Celer cBridge also scores 10 on licence, as does the Apache-licensed code of Chainflip.
- You want one interface that routes over several bridges: Squid settles through its own intents system or hands off to Axelar, Chainflip, Circle's CCTP or LayerZero. Its audit record scores 8, with four named firms and linked reports. It scores 2 on licence, though, because its router contracts are not in any public repository.
Criterion marks side by side
| Bridge | Fees | Trust model | Licence | Chains | Audits | Overall |
|---|---|---|---|---|---|---|
| Axelar | 5 | 9 | 10 | 9 | 10 | 8.6 |
| Wormhole | 5 | 10 | 9 | 9 | 10 | 8.6 |
| Celer cBridge | 7 | 9 | 10 | 9 | 7 | 8.4 |
| deBridge | 9 | 6 | 8 | 7 | 9 | 7.8 |
| Chainflip | 9 | 10 | 10 | 5 | 4 | 7.6 |
| Across | 5 | 10 | 4 | 9 | 8 | 7.2 |
| Stargate / LayerZero | 5 | 8 | 2 | 9 | 9 | 6.6 |
| Squid | 6 | 5 | 2 | 9 | 8 | 6.0 |
| Synapse | 2 | 6 | 9 | 7 | 5 | 5.8 |
How the ranking works
Each bridge gets a 0 to 10 mark on five criteria from the published rubric. On 6 November 2026 every mark was checked against a single scale per criterion, so the same evidence earns the same mark from one bridge to the next.
- Fee transparency: the cost published as figures. A per-chain fee table or an itemized fee schedule scores 9 (deBridge, Chainflip). Celer cBridge gets 7 because its FAQ gives a 0%–2% protocol fee ceiling and its fee section gives 0%–0.5%. Squid gets 6 for a convenience fee on some routes with no published rate. A fee charged on every transfer and explained only as components, with no rate, scores 5 (Axelar, Wormhole, Across, Stargate). A fee that appears only inside each quote scores 2 (Synapse).
- Trust model disclosure: what secures a transfer, in the protocol's own words. A stated quorum or challenge mechanism with its weak points named scores 10. A described mechanism with no validator count or threshold scores 6 (deBridge). A settlement flow with no failure case stated scores 5 (Squid).
- Licence & code openness: read from the repository file. A recognized open-source licence stated in full scores 10. The same licence loses one point when it is a short-form notice (Wormhole) or rests on per-file headers with no LICENSE file (Synapse). A BUSL licence past its change date, where the GPL change licence now applies by its own terms, scores 8 (deBridge). A BUSL licence before its change date scores 4 (Across). A licence with no change licence, one whose text says it is not open source, or no public repository for the core contracts at all scores 2 (Stargate, Squid).
- Chain coverage: networks the protocol lists in its own documentation, with per-product tables counted as a list. Twenty-five or more documented networks score 9. The 18 and 19 listed by deBridge and Synapse score 7, and Chainflip's eight score 5.
- Audit disclosure: reports published with the firms named. A large dated archive from many named firms scores 10 (Axelar, Wormhole). Named, linked reports score 7 to 9 depending on dates and depth: deBridge's 33 dated reports score 9, Across's eight undated reports 8 and Celer's five undated reports 7. Reports that sit in a code repository with no link from the docs score 5 (Synapse). Audits recorded only by the auditor, with no firm named in the protocol's own docs, score 4 (Chainflip).
The overall score is the plain average of the five marks, rounded to one decimal, with equal weight for each. For deBridge that is (9 + 6 + 8 + 7 + 9) / 5 = 7.8. Ties are listed alphabetically, which is why Axelar appears above Wormhole at 8.6. The marks come from desk research of each protocol's documentation, repositories and audit pages, checked on 6 November 2026. We did not send transfers. Marks are re-checked, and each review logs every change with its reason and source.
What changed in this update
The previous edition, checked on 24 July 2026, ranked ten bridges. In this November 2026 update, eight of the remaining nine moved:
- Chainflip, up to 7.6 from 4.6: its governance and security pages returned errors in July. They now state the 100-of-150 vault threshold (trust model 2 to 10); fees, chains, audits and licence also rose.
- deBridge, up to 7.8 from 5.4: its security overview now links a repository of 33 audit reports (audits 2 to 9). The BUSL change date has passed, so licence rose from 4 to 8.
- Synapse, up to 5.8 from 3.4: its documentation now loads. It covers the Intent Network's dispute model (trust 2 to 6), lists 19 networks (chains 2 to 7), and named audit reports sit in its code repository, though the docs do not link them (audits 2 to 5).
- Celer cBridge, up to 8.4 from 6.2: its audit page now loads and links five reports (2 to 7), and its homepage states 40+ chains (3 to 9). Fees fell from 8 to 7 because of the conflicting fee ceilings.
- Wormhole, up to 8.6 from 6.8: per-product network tables raised chain coverage from 3 to 9, and audits rose from 9 to 10.
- Squid, up to 6.0 from 4.8, and Stargate, up to 6.6 from 6.0: Squid gained on audits (2 to 8) but fell on fees (9 to 6). Stargate's own V2 fees page says its flat 6 bps fee applied to V1 pools, which took fees from 9 to 5. Its V2 network list raised chain coverage from 2 to 9.
- Across, down to 7.2 from 7.4: fees went from 6 to 5 under the single fee scale. Axelar was unchanged at 8.6.
Hop Protocol was withdrawn: its bridge app and API are offline, and the hop.exchange domain serves an unrelated site.
What the scores do not cover
- Speed and the price you actually pay: we did not measure either. Settlement times on these pages are the companies' own figures, such as Across's under two seconds, deBridge's 10 to 60 seconds and Celer's 5 to 20 minutes.
- Incident history: the rubric scores incidents only when a protocol documents them itself. Wormhole's Governor design document records the 2 February 2022 exploit of its Solana core contract. "Zero exploits" lines on other sites are those companies' own statements.
- Who runs the protocol: on 15 December 2025, Axelar's blog said Circle had agreed to acquire the Interop Labs team. Across's forum said on 17 September 2026 that the protocol had passed to Across, Inc. Synapse's domain now redirects to an options venue. Celer names no legal entity. None of this changed a mark.
- Liquidity and liveness: pool depth, credit limits on a route and relayer availability can cap or delay a transfer.
- Access: some terms exclude certain users. deBridge's terms, for example, exclude US persons.
- The connected chains: a bridge cannot be safer than the chains at either end of it.
How to choose
- Confirm that both chains and the asset appear in the protocol's own supported-network list, not only in a third-party app.
- Compare the amount you send with the amount quoted to arrive. For bridges that publish no rates, the quote is the only price you get.
- Find out what secures your route. On Stargate the DVN set can differ by route. On Squid it depends on which underlying bridge settles the transfer.
- For a large amount, look for per-chain limits or delays, such as Wormhole's 24-hour Governor delay or Celer's quarantine for larger transfers.
- Check the terms for country restrictions, and use only the official app address.
Our guide to how bridges work and why they are risky explains the validator, optimistic and pool designs in general terms.
Is a higher-ranked bridge safer to use?
Not necessarily. The score measures how much a bridge discloses about fees, security, licence, coverage and audits, not whether its contracts are free of bugs. A clearly documented trust model tells you what you are relying on, but it does not make that assumption hold. An audit is a review at a point in time, not insurance.
How we score this category
A bridge holds your assets while they are in transit, and what secures them in that window is a design choice the protocol either explains or does not. This rubric scores that explanation alongside the cost, the licence and the audit record. We do not score past incidents unless a protocol documents one itself — reporting an exploit from memory, on a page about somebody else's security, is not something we are willing to do.
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 the cost of a transfer is published as figures — a percentage, a flat amount per chain, or an explicit statement that there is no fee — rather than described only as components a user discovers at quote time.
- Trust model disclosure20% of the score
- How clearly the protocol explains what actually secures a transfer: a validator set with a stated quorum, an optimistic challenge window, a light client, or a configurable verifier stack. Naming the assumption, including its weak points, scores far above asserting that the bridge is trustless.
- Licence & code openness20% of the score
- The licence in the repository, read from the file. A recognised open-source licence scores highest; a delayed-open licence scores mid-table; a licence that states in its own text that it is not open source, or no licence file at all, scores lowest.
- Chain coverage20% of the score
- The networks the protocol states it connects, taken from its own documentation. A page that cannot produce a list, or a counter that renders as zero, is scored as the absence of a published answer.
- Audit disclosure20% of the score
- Whether audit reports are published with the firms named. A homepage claiming a number of audits without naming a single auditor is not disclosure.
What does not affect a score
Only the criteria above count. 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.
Frequently asked questions
How were the bridges in this ranking selected?
The ranking covers the cross-chain bridges and routers reviewed on Coin Currents Daily that were operating, with public documentation, when we checked. Each one is scored on the same five criteria. Hop Protocol was withdrawn in November 2026 because its app and API were offline.
How is each bridge's overall score calculated?
Each of the five criteria gets a 0 to 10 mark, and the overall score is their plain average, rounded to one decimal. No criterion carries extra weight. Ties are listed alphabetically, so Axelar sits above Wormhole although both score 8.6.
Which bridge suits a transfer where I want the fee in advance?
deBridge and Chainflip score highest on fee transparency, at 9. deBridge lists a flat fee per source chain plus 4 basis points. Chainflip itemizes its network and pool fees, though deposit gas and destination broadcast costs vary by chain.
What fees or country restrictions should I check before bridging?
Check the gap between what you send and what arrives, any destination gas folded into the source-chain charge, and any frontend fee added by the app you use. Read the terms as well: deBridge's terms, for example, exclude US persons.
How old are the measurements on this page?
All nine bridges were checked on 6 November 2026, from their documentation, repositories and audit pages. Fees, chain lists and licences change, so each review records its own check date and a log of score changes.
Rankings are editorial. Nothing here is financial advice. Editorial policy · How we score · How we're funded.








