Cross-Chain Messaging evaluated across message trust model, application configurable security, forged and stuck message record, and cost per message.
Union
#1 of 7 · published ranking
Union
80ChainChoice Score
4199
Why it leads
Best in the pool on message trust model (9/10; next 6/10)
Cost
Not priced· No comparable price is published
No provider can pay for a position in this table. The code that computes this order cannot read which links earn us a commission, and every build proves it. Every score below rebuilds from the published criteria.Ranking 2026.08 · 4 criteria · 7 products · same inputs, same order
Personal comparison
Cross-Chain Messaging ranked comparison
No comparable price is published; ranking still uses verified product evidence.
RankProviderBest fitScore
1
UnionUnionIBC-compatible messaging that verifies...Top ranked
Message trust model — 9/10Trade-off: The trust model is the strongest in this comparison and the operational maturity is t...
Ranked on 4 published criteria weighted 35/25/25/15, which set goal alignment — 30 of the 86 points. The rest: regional access 20, evidence depth 18, ease of use 10, institutional trust 8. Profile match is shown in breakdowns but carries no weight.
Audit
Infrastructure
Methodology
2026.09.15
published 2026-09-16
Providers tracked
980+
across 119 categories
Last verified
2026-08-11
newest dated pricing or sentiment read
Named on the roster
2 people
managing directors · 6 automated processes
Decision guide
What matters most before choosing in this category
Weighted on who has to be honest for a message to be believed (35), how much of that an application can configure for itself (25), the record of forged and stuck messages (25), and what a message costs (15). Configurability carries a quarter of the weight because in this category the shipped default and the recommended configuration are different things, and the gap is where the money went: one vendor writes that if an application points to the defaults "then you are fully reliant on the LayerZero Labs multisig", and that this path is "intended to be for testing only and should not be used for production applications". Cost carries the least because it does not track security — identical work spans 49.8x across this pool, and the dearest protocol has the smaller verifier set of the two extremes.
What matters most before choosing in this category
Weighted on who has to be honest for a message to be believed (35), how much of that an application can configure for itself (25), the record of forged and stuck messages (25), and what a message costs (15). Configurability carries a quarter of the weight because in this category the shipped default and the recommended configuration are different things, and the gap is where the money went: one vendor writes that if an application points to the defaults "then you are fully reliant on the LayerZero Labs multisig", and that this path is "intended to be for testing only and should not be used for production applications". Cost carries the least because it does not track security — identical work spans 49.8x across this pool, and the dearest protocol has the smaller verifier set of the two extremes.
For a message to be believed by your contract, exactly who has to be honest — and how many of them?
Is that set the shipped default, or something you configured and can change?
When this protocol has failed, did it forge a message or merely stop delivering them?
Current editor lead
Expert review and scoring weights
Union
UnionData checked Aug 2026
IBC-compatible messaging that verifies counterparty consensus with zero-knowledge light clients rather than a signing committee.
IBC-compatible messaging that verifies counterparty consensus with zero-knowledge light clients rather than a signing committee. Strongest on message trust model (9/10): The only design in this set where no named committee has to be honest for a message to be believed. Union describes itself as "a flavor of IBC (and fully IBC compatible), with certain extensions and innovations to make the protocol more efficient, faster, and flexible for builders", with each connection leveraging… Weakest on cost per message (4/10): Explicit limit rather than an estimate: no Ethereum-mainnet quoting function for Union messaging was identified in this research, so the same 32-byte Ethereum-to-Arbitrum-One measurement performed for the other six protocols at block 25,709,399 on 2026-08-08 could not be reproduced for Union. The deployments page… Published price: No per-message price is published or quotable.
Best forMessage trust model — 9/10
Main tradeoffThe trust model is the strongest in this comparison and the operational maturity is the weakest, and the two should not be traded against each other blindly. Mainnet dates to September 2025, the documented public RPC returned 503 when queried twice during this research, and no per-message fee could be quoted from Ethereum — so a light-client design that removes the committee also removes the ability to price and predict the integration the way the incumbent protocols allow. The thirteen mainnet chains were confirmed from Union's own GraphQL endpoint (graphql.union.build/v1/graphql, v2_chains where testnet is false, re-queried during verification and matching this list exactly) rather than from a documentation page.
Verify before signupNo per-message price is published or quotable. The deployments page — where contracts and fee parameters would be enumerated — describes itself only as covering "All deployments of the `ibc-union` stack, light clients, `ucs*` apps, and other miscellaneous contracts deployed by on Union." (docs.union.build/protocol/deployments/, fetched 2026-08-08). Verified during this pass rather than assumed: the fetched page contains zero 0x-prefixed contract addresses. Unlike the other six protocols, no Ethereum-mainnet quoting function for Union messaging was identified, so the 2026-08-08 Ethereum-to-Arbitrum cost is recorded as unverified rather than estimated.
Methodology
How this category is reviewed
Reviewed on message trust model, application configurable security, forged and stuck message record, and cost per message.
The order on this page is the published ranking for this category. Every criterion, weight and source behind it is on the methodology page.
Frequently asked
Questions people ask before choosing messaging
What actually verifies a message here?
A named set of signers in every case except one, and the counts are usually not on the security page. Reading the Ethereum contracts on 2026-08-08: LayerZero's SendUln302 returned four required DVNs for the default config, which its vendor separately describes as a testing path. Hyperlane's defaultIsm() looks nested and is not — a 2-of-2 aggregation whose routing branch for origin 42161 is a 1-of-2 aggregation, and whose two children, a message-id multisig and a merkle-root multisig, both return validatorsAndThreshold() of 3-of-5 over the identical five addresses. Satisfying either branch takes three signatures from the same five keys, so the second branch adds a proof format rather than a second opinion. Axelar publishes no numeric quorum at all, so the concentration had to be computed: of 151 validators returned by its own API, 48 are bonded, and 21 of those reach 60% of square-root-weighted voting power. Quadratic voting does widen the collusion set here — by three entities.
Does the marketing match what the contracts do?
Twice in this pool it does not. Chainlink still sells CCIP on an independent Risk Management Network, while its own architecture documentation states that "the Risk Management Network's automated offchain role is no longer active in current CCIP deployments" — what remains is a manual switch held by one party, invoked when "the CCIP Owner manually initiates a curse". The same page publishes no quorum for the oracle networks doing the actual verifying. Wormhole's Standard Relayer is deprecated and still reports destinations as supported while quoting nothing for them: at Ethereum block 25,709,399 its relayer returned a native price quote of 0 for Arbitrum, Base, BNB Chain, Polygon and Optimism at both 200,000 and 2,000,000 gas, and its default delivery provider returned isChainSupported(23) true with quoteGasPrice(23) and quoteDeliveryOverhead(23) both 0.
Does paying more buy a stronger guarantee?
No, and the spread is large enough to make that worth saying plainly. For a 32-byte payload from Ethereum to Arbitrum with a 200,000 gas limit, quoted at block 25,709,399: Axelar $0.0385, LayerZero $0.0923, Hyperlane $0.0990, Chainlink CCIP $0.5379, deBridge $1.9189 — a 49.8x spread for identical work. The most expensive of those two extremes has the smaller verifier set and the cheapest has one of the largest, so a buyer paying fifty times is not buying fifty times the honest-party requirement. One caveat the research states about itself: Union, the only light-client design here and the strongest trust model in the set, exposed no Ethereum quoting function and its published mainnet RPC returned HTTP 503 on two attempts, so the trust-minimised option is also the one a buyer cannot yet budget.
What did the April 2026 failure actually change?
It priced the trust model. The attack forged verification rather than exploiting a contract, and the vendor accepted fault in writing. Within three weeks the affected issuer had moved its bridge to a competing protocol and a second issuer said it was migrating more than $700 million of tokenised bitcoin infrastructure away from the same vendor. That is why the forged-and-stuck record carries the same weight here as configurability: a protocol that has stopped delivering messages has cost its users time, and a protocol that has delivered a false one has cost them the balance.
Not financial advice · For informational purposes only · Always do your own research
//Analytics consent·GDPR · ePrivacy · TTDSG
ChainChoice measures page views and conversions with two cookieless, EU-hosted services: Plausible and Cloudflare Web Analytics. Nothing loads until you accept, and rankings are identical either way.ChainChoice measures how the engine is used — page views, conversions, referrer — through two cookieless, EU-hosted services: Plausible and Cloudflare Web Analytics. No advertising cookies, no cross-site profile, no data resale. Neither script loads until you accept, and rankings are identical whether you accept or decline.
