Best cross-chain intent infrastructure for swap developers

Combined ranking

1. 1
   Across Protocolacross.to ↗—14 pts

   GPT #2Claude #1Gemini #1

   The most battle-tested intent-based bridging stack — its relayer/settlement model (fills in seconds, optimistic settlement via UMA) has years of mainnet volume, it co-authored ERC-7683 so integrations track the emerging standard, and its SDK/API plus "Across Settlement" let swap developers embed cross-chain fills without running solvers; assumes the developer targets EVM chains, where its coverage and solver depth are strongest.

   + model takes & fixes− hide details

   Claude The most battle-tested intent-based bridging stack — its relayer/settlement model (fills in seconds, optimistic settlement via UMA) has years of mainnet volume, it co-authored ERC-7683 so integrations track the emerging standard, and its SDK/API plus "Across Settlement" let swap developers embed cross-chain fills without running solvers; assumes the developer targets EVM chains, where its coverage and solver depth are strongest.

   Gemini Leads in capital efficiency and fast execution speed (fills in ~2-second average) within EVM L2s. Its co-creation of the ERC-7683 standard ensures a massive, interoperable solver network.

   GPT Excellent execution on supported routes, with fast relayer-funded fills, native assets, strong capital efficiency, ERC-7683 alignment, destination actions, and a clean unified Swap API; near-tied with LI.FI when EVM-centric price and speed matter more than breadth.

   Where it falls short

   per GPT Its roughly two-dozen-chain footprint and EVM emphasis make it less suitable for broad non-EVM coverage, and production API use requires credentials and an integrator ID.

   per Claude EVM-centric — weak fit if you need Solana, Bitcoin, or Cosmos routes, and exotic long-tail tokens can lack relayer liquidity, forcing fallback paths.

   per Gemini Lacks native, non-EVM chain support (like Solana or Bitcoin) without relying on wrapper contracts or external bridge integrations.

2. 2
   deBridgedebridge.com ↗—10 pts

   GPT #3Claude #3Gemini #2

   Its 0-TVL architecture eliminates pool vulnerability risks and slippage. Natively supports Solana-EVM swaps with a flexible developer API. Near-tie with UniswapX on EVM, but ranked higher for developers due to native Solana/non-EVM support.

   + model takes & fixes− hide details

   Gemini Its 0-TVL architecture eliminates pool vulnerability risks and slippage. Natively supports Solana-EVM swaps with a flexible developer API. Near-tie with UniswapX on EVM, but ranked higher for developers due to native Solana/non-EVM support.

   GPT Proven intent-based liquidity network with fast non-custodial cross-chain and same-chain swaps, broad heterogeneous-chain support, flexible API/widget integration, transaction construction, order tracking, hooks, and integrator monetization.

   Claude Zero-TVL intent/order-flow design with consistently fast fills, native Solana↔EVM support that most intent rivals lack, and a clean API for programmatic order creation — the strongest pick when non-EVM reach matters; near-tie with Relay for this slot.

   Where it falls short

   per GPT Its security and settlement design is more dependent on deBridge’s validation infrastructure than developers seeking maximally neutral or modular verification may accept.

   per Claude Smaller solver/market-maker set than Across means wide or unusual orders can see worse pricing or slower fills, and you're trusting a more concentrated filler network.

   per Gemini Fills depend entirely on active solver capital on destination chains, leading to high fees or delayed fills on less popular token routes.

3. 3
   LI.FIli.fi ↗—10 pts

   GPT #1Claude #2Gemini #5

   The strongest default for typical swap developers: one mature SDK/API integration spans 60+ chains, aggregates bridges, DEXs, and intent routes, supports destination calls and integrator fees, while LI.FI Intents adds fast solver-funded execution and permissionless Open Intents Framework compatibility.

   + model takes & fixes− hide details

   GPT The strongest default for typical swap developers: one mature SDK/API integration spans 60+ chains, aggregates bridges, DEXs, and intent routes, supports destination calls and integrator fees, while LI.FI Intents adds fast solver-funded execution and permissionless Open Intents Framework compatibility.

   Claude Best single integration surface for a swap product: aggregates intent systems (Across, Relay, deBridge) alongside classic bridges and DEXes with routing, contract calls, and solid enterprise support/SLAs, so developers get intent-speed routes where available and fallbacks everywhere else; ranked on breadth-plus-reliability rather than being pure intent infra itself.

   Gemini Meta-aggregator providing a single integration point for both intent-based bridges and classic routing, automatically directing order flow to the best path.

   Where it falls short

   per GPT Its aggregation and hosted routing layers add dependency and complexity; teams wanting the smallest trust surface or direct protocol control should integrate a settlement protocol instead.

   per Claude It's an aggregation layer with fee overhead and less control — you inherit its routing decisions and dependency surface rather than owning the intent flow directly.

   per Gemini Adds an extra layer of smart contract risk and fee overhead, while locking developers into LI.FI's centralized routing API.

4. 4
   UUniswapX—3 pts

   GPT —Claude —Gemini #3

   Combines solver auctions with Uniswap's deep v2/v3/v4 liquidity pools for optimal pricing. Near-tie with deBridge, but ranked lower due to lack of non-EVM support and Uniswap-centric architecture.

   + model takes & fixes− hide details

   Gemini Combines solver auctions with Uniswap's deep v2/v3/v4 liquidity pools for optimal pricing. Near-tie with deBridge, but ranked lower due to lack of non-EVM support and Uniswap-centric architecture.

   Where it falls short

   per Gemini Heavily integrated into Uniswap's routing and license ecosystem, making it restrictive and less modular for independent developers who want a general-purpose cross-chain SDK.

5. 5
   MMayan—2 pts

   GPT #4Claude —Gemini —

   Particularly strong for Solana-to-EVM swaps: competitive driver auctions provide guaranteed output and near-instant inventory-funded delivery, while its SDK, gas-drop support, payload execution, referral fees, and low 3-bps protocol fee make integration practical.

   + model takes & fixes− hide details

   GPT Particularly strong for Solana-to-EVM swaps: competitive driver auctions provide guaranteed output and near-instant inventory-funded delivery, while its SDK, gas-drop support, payload execution, referral fees, and low 3-bps protocol fee make integration practical.

   Where it falls short

   per GPT Its chain footprint is narrower and its Wormhole/Solana-centered auction and settlement architecture is a poor fit for teams needing chain-neutral infrastructure or extensive long-tail coverage.

6. 6
   NNEAR Intents—2 pts

   GPT —Claude —Gemini #4

   Features native cross-chain chain-abstraction across EVM, Solana, NEAR, and Bitcoin. Solvers can tap CeFi liquidity, enabling competitive pricing on large-size transactions.

   + model takes & fixes− hide details

   Gemini Features native cross-chain chain-abstraction across EVM, Solana, NEAR, and Bitcoin. Solvers can tap CeFi liquidity, enabling competitive pricing on large-size transactions.

   Where it falls short

   per Gemini Settles transactions on the NEAR blockchain, forcing developers to inherit NEAR's trust assumptions and infrastructure dependencies.

7. 7
   RRelay—2 pts

   GPT —Claude #4Gemini —

   Extremely low-cost, near-instant fills across a very long list of chains including many L2s/appchains others skip, with a simple API that made it the default for bridging-into-app UX; earns the spot on latency, cost, and chain coverage for small-to-mid-size swaps.

   + model takes & fixes− hide details

   Claude Extremely low-cost, near-instant fills across a very long list of chains including many L2s/appchains others skip, with a simple API that made it the default for bridging-into-app UX; earns the spot on latency, cost, and chain coverage for small-to-mid-size swaps.

   Where it falls short

   per Claude Relies heavily on Relay's own filler capacity and reputation-based settlement — larger trades and trust-minimization-sensitive use cases fit worse than Across's model.

8. 8
   11inch Fusion+—1 pts

   GPT #5Claude —Gemini —

   A strong choice for gasless, MEV-resistant cross-chain swaps using competitive resolvers and atomic escrow settlement, backed by excellent same-chain liquidity aggregation and polished developer APIs.

   + model takes & fixes− hide details

   GPT A strong choice for gasless, MEV-resistant cross-chain swaps using competitive resolvers and atomic escrow settlement, backed by excellent same-chain liquidity aggregation and polished developer APIs.

   Where it falls short

   per GPT Fusion+ is comparatively constrained by supported routes, resolver participation, escrow timing, and commercial API terms, so it is less universal as an independent cross-chain backend.

9. 9
   BBungee—1 pts

   GPT —Claude #5Gemini —

   Socket's intent protocol plus its established aggregation network give developers auto-routed intents with solver competition and a mature API used by major wallets; solid but ranks last because its intent layer is younger and solver depth thinner than the picks above — near-tie with an aggregator-shaped LI.FI alternative role.

   + model takes & fixes− hide details

   Claude Socket's intent protocol plus its established aggregation network give developers auto-routed intents with solver competition and a mature API used by major wallets; solid but ranks last because its intent layer is younger and solver depth thinner than the picks above — near-tie with an aggregator-shaped LI.FI alternative role.

   Where it falls short

   per Claude Overlaps heavily with LI.FI as an aggregation surface; if you already integrate LI.FI or Across directly, Bungee adds little unique capability.

Just missed the top 5

By model

Common questions

Tracked by ModelsAgree · rank 1 = 5 pts … rank 5 = 1 pt · re-polled weekly