The "best L2 bridge" is rarely one bridge. Every major Layer 2 ships a native rollup bridge that is canonical, has zero protocol fee, and is slow on the withdrawal leg. Next to it sit third-party fast paths: Circle's CCTP for native USDC, Hyperlane warp routes for ERC-20s, intent-based bridges and aggregators for everything else. The right route depends on direction, asset, size, and patience.
​

This guide covers nine L2s: Arbitrum, Optimism, Base, Polygon zkEVM, zkSync Era, Linea, Scroll, Mantle, and Blast. Each has the same shape, with finality that differs by rollup type.
​

An L2 bridge moves assets between Ethereum L1 and an L2 rollup, or between two L2s. Every rollup has a canonical "native" bridge written into the chain's settlement contracts. Third-party bridges sit on top, using messaging layers (CCTP, Hyperlane, LayerZero) or solver networks (Across) to deliver faster paths that bypass the rollup's withdrawal delay.
​

The native bridge is the trust-minimized path. Deposits typically confirm in minutes because they only require an L1 transaction and an L2 inclusion. Withdrawals are where the asymmetry shows up. Optimistic rollups enforce a challenge window so anyone can dispute fraud. ZK rollups wait for a validity proof to verify on L1. Both processes take time, and that gap is exactly what fast-path bridges sell against.
​

All L2 bridges follow the same shape because their security depends on Ethereum L1 finality. The native bridge inherits the rollup's full security model and enforces the rollup's exit timing. Fast bridges sit outside that timing by fronting destination liquidity through a separate trust assumption, usually a relayer network, a messaging layer, or a mint-and-burn protocol.
​

On optimistic rollups (Arbitrum, Optimism, Base, Mantle, Blast), withdrawals carry a 7-day challenge window during which a validator can submit a fraud proof. Per L2BEAT data, Blast reduced its window from 14 days to 7 days in 2024, matching the OP Stack default. Mantle adds a ~60-minute prove step before the 7-day window starts.
​

ZK rollups skip the challenge step. Linea, Scroll, zkSync Era, and Polygon zkEVM use validity proofs that confirm correctness cryptographically once verified on L1. Withdrawals range from roughly 30 minutes to several hours depending on proof cadence. Linea's docs show ZK proof finalization for L1-bound withdrawals takes hours under normal load and is targeted to shrink further through 2026 protocol upgrades.
​

Fast-path bridges side-step both windows. Across uses solvers and an optimistic oracle to front destination liquidity in seconds. CCTP burns USDC on source and mints on destination through a Circle-attested message, also in seconds. Hyperlane warp routes lock-and-mint or burn-and-mint within minutes through a configurable messaging layer. Each fast path trades native-bridge security for speed.
​

The table below summarizes the native bridge plus the dominant fast paths for the nine L2s in this guide. Native bridge fees are zero in every case; the cost is L1 gas plus the withdrawal wait. Fast-path fees vary by route, asset, and size, and the underlying transport (CCTP, Hyperlane, intent solvers) is reflected per column.
​

CCTP coverage is the cleanest split. Per the Circle CCTP registry, native USDC mint-and-burn is live on Arbitrum, OP Mainnet, Base, Polygon PoS, Linea, plus an expanding list. Coverage is not yet live for zkSync Era, Polygon zkEVM, Scroll, Mantle, or Blast as of Q1 2026. Hyperlane warp routes have the widest L2 coverage by chain count, with the Hyperlane registry showing 200+ active routes across 140+ chains.
​

Native L2 bridges lock-and-mint on deposits and burn-and-release on withdrawals. The deposit transaction calls an L1 contract that escrows the asset and emits a message; the L2 sequencer picks up the message and credits the user on L2 within minutes. Withdrawals reverse the flow, gated by the rollup's exit rules.
​

For the four major optimistic rollups in this list (Arbitrum, OP Mainnet, Base, Blast) the exit takes ~7 days. Arbitrum has shipped BOLD, a permissionless fraud-proof system, so any validator can dispute a malicious withdrawal. Optimism, Base, and Blast are migrating to permissionless fault proofs along the OP Stack rollout plan; check the L2BEAT risk dashboard for current status. Mantle layers a ~60-minute prove step on top of the 7-day window.
​

ZK rollups skip the challenge entirely. Polygon zkEVM posts validity proofs roughly every 30 to 60 minutes, so a withdrawal becomes claimable once the batch containing it is proven on L1. Scroll finalizes most withdrawals in 3 to 5 hours. Linea and zkSync Era cluster in the same range, with Linea's L1 hard finality currently in the multi-hour band and on track to compress through Q1-Q2 2026 protocol updates per Linea's public roadmap.
​

Circle's Cross-Chain Transfer Protocol burns USDC on the source chain and mints native USDC on the destination through a Circle-attested message. No wrapped representation, no liquidity pool, no slippage. CCTP V2 became canonical when V1 was deprecated in 2025, and adds Fast Transfer with sub-minute attestation on V2-enabled chains.
​

For L2s where CCTP is live, it is the dominant route for any meaningful USDC movement, per the Circle CCTP overview. The user pays source gas, destination gas, and a small per-transfer fee that depends on Fast Transfer settings. Where CCTP is not yet live, USDC moves either through bridged-USDC representations (USDC.e) backed by the native bridge, or through an aggregator that hops through a CCTP-supported chain on the way.
​

Hyperlane warp routes carry the ERC-20 footprint that USDC alone cannot cover. A warp route is a deployed pair of contracts (collateral lock on source, synthetic mint on destination) governed by a configurable Interchain Security Module that the route's operators choose at deployment. The result is permissionless token bridging across any Hyperlane-connected pair.
​

MNT moves between Ethereum and Mantle via Hyperlane warp routes. Blast's yield-bearing assets bridge through Hyperlane's Yield Routes, which preserve source-chain yield on the destination representation. Tether's USDT0 uses LayerZero's OFT standard alongside Hyperlane for messaging across L2s. For L2s without CCTP coverage, Hyperlane warp routes are often the cleanest non-aggregator path for non-USDC ERC-20s.
​

Intent-based bridges and aggregators are the fast-path layer above CCTP, Hyperlane, and native bridges. An intent-based bridge takes a user's signed "I want X on chain Y" message and lets solvers compete to fill it. The solver fronts destination liquidity in seconds, then reclaims from the canonical bridge on its own timeline.
​

Across is the original intent-based bridge and a co-author of ERC-7683, the cross-chain intent standard now backed by Arbitrum, OP Mainnet, Polygon, zkSync, Linea, Scroll, and others. Across is one of the fastest paths most commonly used between Ethereum L1 and major L2s for USDC, ETH, and WBTC.
​

Aggregators (LI.FI, Squid, Jumper, deBridge, Socket, Rango) sit one layer above. They route a transfer through whichever bridge or DEX combination quotes the best price, composing existing bridges and DEX swaps into a single user-signed transaction. Eco Routes is an intent-based router on top of these rails, exposing a unified routing API across CCTP, Hyperlane, LayerZero, and the intent-based bridges underneath.
​

Every major L2 here still runs a centralized sequencer as of Q1 2026. A single entity (Offchain Labs, OP Labs and Conduit, Matter Labs, ConsenSys, the Scroll team, Mantle, Blast) controls L2 transaction ordering. This matters for bridge UX because a sequencer outage can stall both deposits and L2-side withdrawal proofs.
​

The path forward is in progress. Espresso Systems ships a shared sequencer that the OP Superchain plans to adopt across member chains alongside the native interop launch. Arbitrum has decentralized validation through BOLD; sequencer decentralization is on its 2026-2027 roadmap. zkSync targets multi-party sequencing with shared revenue in the same window per public roadmap statements.
​

For bridge users today, deposits depend on the sequencer including the transaction, and L1-side fault proofs depend on the sequencer publishing state. Native bridges have force-inclusion mechanisms as a backstop if the sequencer censors. Fast-path bridges with their own relayer networks (Across, Hyperlane, CCTP) bypass sequencer discretion, which is part of why they exist.
​

For native bridges, ZK rollups finalize fastest. Polygon zkEVM withdrawals clear in ~30 to 60 minutes; Scroll in 3 to 5 hours; Linea and zkSync Era in roughly that range. For fast-path bridges, intent-based bridges like Across and CCTP V2 Fast Transfer settle in seconds on L2s where they are live.
​

Yes, through any fast-path bridge or aggregator. CCTP supports direct L2-to-L2 USDC moves. Hyperlane warp routes connect any two Hyperlane-supported chains directly. Aggregators like LI.FI, Squid, and Jumper compose direct L2-to-L2 routes for most token pairs. Native rollup bridges only connect to L1; cross-L2 via native bridges requires two hops.
​

Circle adds CCTP support chain by chain after integration and audits. As of Q1 2026, CCTP V2 is live on the most established L2s (Arbitrum, OP Mainnet, Base, Polygon PoS, Linea, plus an expanding set) and not yet on Polygon zkEVM, zkSync Era, Scroll, Mantle, or Blast. The Circle developer docs are the source of truth.
​

Sources and methodology. Native bridge withdrawal times, sequencer status, and ZK proof cadence verified against L2BEAT, official protocol documentation (Linea, Scroll, zkSync, Blast, Mantle), and the Circle CCTP developer registry. Hyperlane chain coverage from the public Hyperlane documentation. Figures reflect Q1 2026 state and refresh as bridges, sequencers, and CCTP coverage expand.
​