Ethereum Layer 2s: A Research Guide

Quick answer: The best Ethereum L2 in 2026 depends on your use case. Arbitrum leads in DeFi TVL (~$19B, Stage 1 with BoLD permissionless fraud proofs). Base leads in daily transaction volume (12.89M/day), though it departed the OP Stack in February 2026. Optimism anchors the 34-chain Superchain. For ZK technology with Stage 1 security, choose Scroll (EVM-equivalent, Stage 1) or zkSync Era (native account abstraction, Stage 0). Linea (ConsenSys/MetaMask, Stage 0) launched its LINEA token in September 2025. Post-Fusaka (December 2025), all major L2 swaps cost under $0.02 — fees are not the differentiator.

State of Ethereum L2s in 2026

Arbitrum owns DeFi TVL, Base owns daily transaction volume, and the Superchain owns multi-chain governance — but the map shifted in February 2026 when Base departed the OP Stack. Post-Fusaka, fees are negligible everywhere; what separates chains now is proof-system maturity, ecosystem depth, and sequencer trust. Last verified: 2026-05-27.

73+ active rollups secure over $48B in combined L2 TVL as of May 2026. Arbitrum One ($19B TVL, per L2BEAT) and Base ($12B TVL) together account for roughly 65–70% of all L2 liquidity. The long tail of 50+ rollups launched in 2023–2025 mostly followed the same trajectory: incentivize TVL with token programs, watch deposits land, watch 70–90% leave within weeks of program expiry. Manta, Mode, Linea's first incentive season, Blast at peak — same chart shape every time.

Three things separated the survivors from the failures. Native distribution (Base has Coinbase's retail onboarding funnel; Arbitrum had a four-year head start with a real DeFi user base). Sticky non-farming use cases (Farcaster on Base, GMX on Arbitrum, Velodrome on Optimism). And proof-system credibility — the post-EIP-4844 blob market equalized DA costs, but the rollups with audited proof systems and Stage 1 certification attracted institutional deposits that new chains couldn't.

A critical upgrade: Ethereum's Fusaka hard fork activated on December 3, 2025, deploying PeerDAS (EIP-7594) and the first BPO blob parameter forks. Blob target and maximum jumped from 6/9 (set by Pectra in May 2025) toward 10/15 and eventually 14/21 per block. L2 DA costs fell an estimated 40–60% within the first month. The effect on mainnet Ethereum gas was also sharp — average Ethereum transaction fees dropped toward $0.01 post-Fusaka. L2 fees, already sub-cent on most chains, compressed further.

Optimistic vs ZK — the architectural fork

• Optimistic rollups (Arbitrum, Optimism, Base, Blast): assume transactions are valid; allow a 7-day window for fraud proofs. Long withdrawal latency to L1; bypass via third-party bridges.
• ZK rollups (ZK rollup = zero-knowledge rollup, batches transactions with a cryptographic validity proof posted to L1) (zkSync Era, Scroll, Linea, Starknet, Polygon zkEVM): submit cryptographic validity proofs with every batch; no challenge period, fast L1 finality once proof lands.

The honest take in 2026: ZK is technically superior on every axis that matters (faster finality, no 7-day withdrawal, smaller DA footprint) and proof costs are converging fast with Fusaka. ZK still trails on tooling maturity and EVM corner cases — though Scroll's bytecode equivalence and Linea's Type-1 zkEVM launch in early 2026 have closed the gap significantly. Most new launches in 2026 are ZK; new optimistic launches are rare. The optimistic incumbents still hold the majority of TVL.

L2BEAT Stage classification — what it means

L2BEAT's Stage 0 / Stage 1 / Stage 2 framework grades rollup maturity on proof-system decentralization and user exit rights:

• Stage 0 — training-wheels rollup. Operator multisig can pause or upgrade contracts without a time-lock. Users must trust the operator to not steal funds.
• Stage 1 — fraud proofs or validity proofs are live; users can exit to L1 without operator cooperation; contract upgrades require a minimum 7-day notice window. Security council can still intervene in extreme cases.
• Stage 2 — fully permissionless proofs, no security council override outside narrow on-chain-defined recovery scenarios. No major rollup has reached Stage 2 as of May 2026.

Current stages (L2BEAT, May 2026): Arbitrum One (Stage 1, BoLD), Base (Stage 1, OPFP), OP Mainnet (Stage 1, OPFP), Scroll (Stage 1), Starknet (Stage 1, Stwo). zkSync Era (Stage 0), Linea (Stage 0), Blast (Stage 0).

Top 7 Ethereum L2s in 2026

Arbitrum (DeFi depth, Stage 1), Base (consumer apps, 12.89M daily txns), Optimism (Superchain anchor, Stage 1), Scroll (Stage 1 ZK), zkSync Era (Stage 0 ZK, native AA), Linea (Stage 0 ZK, MetaMask), Blast (Stage 0, yield-bearing). Last verified: 2026-05-27.

TVL figures from L2BEAT (May 2026). Daily txn figures from growthepie/L2BEAT. Swap cost estimates post-Fusaka.

Arbitrum — deep dive

Arbitrum is where the real DeFi money lives. It is the L2 every serious DeFi protocol launches on first and where professional market makers size their largest positions.

Best for

DeFi power users, institutional liquidity, GMX / Aave / Pendle / Uniswap v4 trading. When Wintermute or Flow Traders quote a size on-chain, they quote it on Arbitrum first.

Why it leads

Arbitrum shipped first with full EVM equivalence, survived three protocol upgrades (Nitro → Stylus → BoLD) without a chain-halting exploit, and has 1.97B+ total transactions processed since launch. BoLD (Bounded Liquidity Delay) went live on mainnet, replacing the permissioned validator whitelist with fully permissionless fraud proofs — anyone can run a validator node and challenge invalid state transitions. This pushed Arbitrum to Stage 1 and unlocked institutional-grade trust assumptions. Offchain Labs retains a Security Council multisig (preventing Stage 2), but the fraud-proof layer is now open.

Trade-offs

The sequencer is centralized (operated by Offchain Labs) — single point of failure for censorship, not for funds, since users can force-include transactions via the L1 bridge. The 7-day L1 withdrawal challenge period is real — bypass via Across or CCTP for normal use. Stylus (Rust/C++ contracts) remains technically impressive but adoption has been slower than hoped.

TVL and throughput

~$19B TVL (~40–44% of all L2 market share), ~$4.2B stablecoin reserves, ~62 TPS sustained, 4.3M daily transactions (L2BEAT / eco.com, May 2026). Arbitrum has held the top-TVL L2 position since mid-2023; the gap vs. competitors is widening, not closing.

Fees

Under $0.02 per Uniswap v3 swap post-Fusaka. Stylus contracts can run 10–100x cheaper than equivalent Solidity for compute-heavy workloads.

Base — deep dive

Base has a direct distribution channel into Coinbase's 100M+ verified users — no other L2 has a comparable fiat-to-chain onramp.

Best for

Consumer apps, Farcaster Frames, social-fi, AI-agent apps, meme coins, and onboarding non-crypto users via Coinbase Smart Wallet (gasless first transaction). If you're building for an audience that doesn't already own a hardware wallet, launch on Base.

The February 2026 OP Stack departure

On February 18, 2026, Coinbase announced Base would migrate off the OP Stack to a proprietary "unified stack" built on Reth (Rust-based Ethereum client by Paradigm). The stated goals: six hard forks per year (vs. three under OP Stack), full control over sequencer revenue (Base previously shared revenue with the Optimism Collective under a 2023 agreement), and faster iteration. The new stack retains ~99% of the underlying code and the same optimistic rollup mechanics with a 7-day challenge window. Base's L2BEAT Stage 1 classification was maintained. The OP token fell ~7% on the news.

Why it leads in users

Coinbase Smart Wallet went mainstream in 2024–2025 as a passkey-secured account-abstraction wallet requiring no seed phrase. Coupled with Base's per-user free transaction allowance for new accounts, onboarding time dropped from ~20 minutes to ~30 seconds. Farcaster's (Farcaster is a decentralized social protocol with most activity on Base) growth, Zora creator economy, and the AI-agent app cohort (Virtuals, Clanker, Higher) all concentrate on Base.

Trade-offs

Coinbase-operated sequencer creates regulatory exposure — if Coinbase faces legal pressure, Base inherits it directly. The February 2026 stack migration raised concerns about long-term alignment with the broader Ethereum rollup roadmap. No native Base token, so no airdrop farming effect — cleaner user base but lower speculator-driven TVL ceiling.

TVL and users

~$12B TVL, 12.89M daily transactions, ~382,500 daily active users (L2BEAT / eco.com, Q1–Q2 2026). TVL grew from $3.1B in Jan 2024 to $11.2B by late 2025 and stabilized near $12B in 2026. Highest daily transaction count of any L2 by a wide margin.

Fees

Under $0.02 per swap post-Fusaka. Coinbase Smart Wallet provides gasless first transaction for new users.

Optimism — deep dive

Optimism is the platform play. OP Mainnet itself is smaller than Arbitrum or Base; the OP Stack — open-source rollup framework — was the real product, powering Base, World Chain, Mode, Unichain, Soneium, and 34 chains in total.

The Superchain without Base

Base's February 2026 departure materially changed the Superchain narrative. Before February 2026, Base was the Superchain's largest chain by TVL by a wide margin. Post-departure, the remaining 33 publicly announced Superchains hold approximately $6B combined TVL (L2BEAT / Messari H1 2025 baseline). Unichain contributes ~$880M TVL (dominated by Uniswap v4). World Chain contributes ~$411M. OP Mainnet itself holds ~$1.5B TVL. The Superchain still processes 17.3M+ daily transactions across its remaining chains.

Stage 1 and fault proofs

OP Mainnet achieved Stage 1 with the launch of permissionless fault proofs (OPFP). Any user can challenge an invalid state transition; withdrawals of ETH and ERC-20s from OP Mainnet to L1 require no trusted third party.

Trade-offs

OP Mainnet's standalone TVL is modest relative to Arbitrum or Base. The Superchain interop upgrade — atomic cross-chain composability across all OP Stack chains — remains in devnet/testnet as of May 2026, with mainnet deployment targeting the Pectra-aligned upgrade window later in 2026. The value of OP-the-token depends significantly on this interop work landing and sequencer revenue pooling across the ecosystem.

TVL and users

~$1.5B TVL (L2BEAT), 2.35M daily transactions, ~19,300 daily active users (eco.com, early 2026).

Fees

Under $0.02 per swap.

Scroll — deep dive

Scroll is the most EVM-faithful ZK rollup in production, and it graduated to Stage 1 — placing it ahead of the larger zkSync Era and Linea by trust-minimization.

Best for

Protocols migrating audited Solidity contracts from Ethereum mainnet with zero rewrite. EVM bytecode deploys unchanged. Ideal for DeFi teams that need ZK finality without audit surface changes.

Proof system and Stage 1

Scroll posts cryptographic validity proofs to L1, giving sub-hour finality with no challenge period. Its Stage 1 classification confirms: users can exit funds to Ethereum L1 without operator cooperation, and upgrades require a minimum 7-day notice. TVL is approximately $2.1B, making Scroll the largest ZK rollup by TVL that has also achieved Stage 1.

Trade-offs

Smaller ecosystem and weaker distribution channels than Linea (no MetaMask integration) or zkSync Era. Growth has been steady but not explosive — Scroll relies on developer trust and technical merit rather than token-incentive programs.

zkSync Era — deep dive

zkSync Era has the most ambitious proving infrastructure of any L2 — and the most centralized control, per L2BEAT's Stage 0 rating.

Best for

Apps that lean hard on account abstraction UX — every account is a smart contract by default, gas payable in any token (ERC-20 fee abstraction native). Best developer-experience ZK rollup for AA-native applications.

Airbender prover and ZK token

In early 2026, Matter Labs shipped Airbender, a RISC-V zkVM prover reaching 21.8 MHz on a single H100 GPU — over 6x faster than competing zkVMs. The ZK token launched via airdrop in June 2024. In May 2026, the V31 upgrade introduced native interoperability between ZKsync chains, with transactions denominated in ZK. zkSync Lite deprecation is planned for 2026.

Stage 0 and trust assumptions

L2BEAT classifies zkSync Era as Stage 0: a multisig council retains unilateral ability to pause or upgrade contracts. Matter Labs has published a decentralization roadmap toward Stage 1 but has not confirmed a production timeline as of May 2026. Note: zkSync Era's proof system was briefly manually paused due to a disclosed vulnerability — a precedent that Stage 0 classification appropriately captures.

TVL

~$404M TVL (L2BEAT, May 2026) — a significant decline from peak but stabilized.

Linea — deep dive

Linea is the ConsenSys bet: MetaMask distribution plus a ZK proof system, now with a live token and a Type-1 zkEVM.

Notable 2025–2026 developments

• LINEA token launched September 10, 2025 — 72B total supply, 85% community/ecosystem allocation, 15% to ConsenSys locked 5 years, no private investor allocation. Dual burn: 20% of net ETH profits burned as ETH, 80% used to buy back and burn LINEA.
• Type-1 zkEVM shipped in early 2026: any Ethereum dApp deploys on Linea without code modification, the strongest EVM-equivalence guarantee.
• Lineth (Linea's ZK rollup stack) joined the Linux Foundation Decentralized Trust in May 2026 as a neutral open-source project.
• SWIFT tested cross-border payments on Linea (reported May 2026).
• Aave restored WETH borrowing on Linea in May 2026 after earlier restrictions.

Stage 0 trust assumptions

Linea remains Stage 0 per L2BEAT. ConsenSys controls both the sequencer and the prover; bridge contracts are upgradeable by a multisig without a time-lock. Higher operator trust dependency than Stage 1 chains.

TVL

~$440M TVL (L2BEAT, May 2026). The first incentive-season TVL collapse in 2024 was real; the team has rebuilt with LXP loyalty points, the token launch, and the Type-1 upgrade.

Blast — the yield rollup

Blast remains a Stage 0 optimistic rollup with no live fraud proofs and declining on-chain revenue in 2026. Native ETH yield (via Lido restaking) and stablecoin yield (USDB, via T-bill exposure) accrue by default to all balances — the "idle balances earn" pitch. The trade-off: the ecosystem skews toward higher-risk DeFi, and without fraud proofs, users rely entirely on the Blast multisig not acting adversarially. Blast's bridge launched as a 3-of-5 multisig controlled by Blast contributors; as of May 2026, no permissionless fraud-proof system has shipped. L2BEAT classifies Blast as Stage 0. Network activity generated negative on-chain revenue in March 2026, reflecting weak sustained usage beyond the initial incentive period.

Best L2 by use case

DeFi liquidity: Arbitrum. Consumer/onboarding: Base. ZK tech + Stage 1: Scroll. ZK tech + AA: zkSync Era. Yield-on-balance: Blast (Stage 0 risk). Superchain: Optimism. Last verified: 2026-05-27.

• Best L2 for DeFi liquidity — Arbitrum (~$19B TVL, GMX, Aave, Pendle, Uniswap v4, deepest order books).
• Best L2 for consumer apps and social — Base (Farcaster, Zora, Coinbase Smart Wallet, 12.89M daily txns).
• Best L2 for new users — Base via Coinbase Smart Wallet (gasless first transaction, passkey login).
• Best L2 for ZK + Stage 1 security — Scroll (bytecode EVM equivalence, Stage 1, ~$2.1B TVL).
• Best L2 for ZK + account abstraction — zkSync Era (native AA, Airbender prover, Stage 0 — trust trade-off applies).
• Best L2 for MetaMask users — Linea (ConsenSys-backed, LINEA token, Type-1 zkEVM, Stage 0).
• Best L2 for yield-on-balance — Blast (native ~4% ETH / ~5% stables — Stage 0 risk, no fraud proofs).
• Best L2 for Superchain ecosystem — Optimism (34-chain anchor post-Base departure, Stage 1).
• Best L2 for institutional treasury — Arbitrum (deepest stablecoin liquidity, Stage 1, mature audit record).
• Best L2 to avoid — Any rollup whose TVL collapsed 70%+ post-incentive expiry (long tail of 50+ failed rollups).

Choosing an L2 by use case

DeFi and large positions: Arbitrum. Consumer and social: Base. Superchain governance: Optimism. ZK with highest trust: Scroll. ZK with AA: zkSync Era. Yield-bearing balances: Blast (with eyes open on Stage 0 risk). Last verified: 2026-05-27.

DeFi (large positions, deep liquidity)

Arbitrum. Most TVL, most protocols, most stable infrastructure. GMX, Aave, Uniswap, Pendle all have the deepest on-chain markets. Stage 1 with BoLD permissionless fraud proofs.

Consumer apps, memes, social

Base. Farcaster Frames, Zora, AI-agent apps, and Coinbase-backed onboarding. Smart Wallet (gasless, passkey-secured) makes onboarding non-crypto users actually feasible. Highest daily transaction count of any L2.

Superchain ecosystem

Optimism (OP Mainnet). Lower standalone TVL than Arbitrum but cleaner governance, Velodrome DEX, Stage 1 fault proofs, and a 34-chain ecosystem after Base's departure. Cross-chain interop targeting late 2026.

ZK technology with Stage 1 trust

Scroll. Bytecode EVM equivalence, validity proofs, Stage 1 certification. Deploy existing Solidity contracts without modification; exit to L1 without operator cooperation.

ZK technology with account abstraction

zkSync Era. Native AA means every account is a smart contract — gas payable in any ERC-20. Stage 0, so higher trust trade-off. Airbender prover is the fastest ZK proving infrastructure in production.

Yield-bearing L2

Blast. Native yield on ETH and stables means idle balances earn. Stage 0 with no fraud proofs — a real trust assumption. Sustained activity metrics are weak post-incentive period.

L2 fees in detail

Post-Fusaka (December 2025), DA costs fell 40–60%. Uniswap v3 swaps cost under $0.02 on Arbitrum/Base/Optimism, under $0.01 on Scroll/zkSync/Linea. Ethereum mainnet costs $0.01–$5 depending on complexity. Last verified: 2026-05-27.

The Ethereum DA cost stack has gone through three compression events:

1. EIP-4844 (March 2024) — introduced blob transactions, cut L2 DA costs roughly 10x versus calldata.
2. Pectra (May 7, 2025) — EIP-7691 doubled blob throughput: target 3→6, max 6→9 blobs per block.
3. Fusaka (December 3, 2025) — PeerDAS (EIP-7594) plus BPO forks raised blob target/max to 10/15 (BPO1) and ultimately 14/21 (BPO2). Estimated 40–60% additional L2 DA cost reduction in first month. Nodes sample only ~1/8 of each blob, enabling far higher throughput without proportional hardware increases.

Post-Fusaka, L2 costs break down as:

1. L1 DA fee (~70–80%) — now dramatically lower with PeerDAS blob sampling.
2. L2 sequencer fee (~20–30%) — flat, typically under $0.001.

Typical costs per Uniswap v3 swap (May 2026):

Insight: EVM L2s have eliminated gas as a meaningful cost for retail after three DA compression events. The fee floor in 2026 is DA cost, not network congestion. Solana is still cheaper but Ethereum L2s have closed the UX gap to within an order of magnitude.

Bridging to L2s

Official bridge for first deposits (gas-only); Across or Stargate for L2-to-L2 (roughly 30 seconds, ~0.05% fee); CCTP for native USDC across all major L2s at zero protocol fee. Last verified: 2026-05-27.

Cheapest: Use the official bridge (Arbitrum Bridge, Base Bridge, etc.) for initial deposits — gas only, no protocol fee.

Fastest: Across or Stargate for L2-to-L2 transfers. Roughly 30 seconds, ~0.05% fee. Skips the 7-day challenge period for optimistic rollup withdrawals entirely.

For USDC: Always use CCTP — native USDC, zero protocol fee, available on all major L2s.

See Best Crypto Bridges 2026 for full comparison.

Security model summary

Stage 1 chains (Arbitrum, Base, OP Mainnet, Scroll) give users permissionless exit rights and upgrade notice periods. Stage 0 chains (zkSync Era, Linea, Blast) do not. All major L2s still run centralized sequencers. Last verified: 2026-05-27.

All major L2s currently have:

• Smart contract bridge on Ethereum L1 (multiply audited)
• Centralized sequencer (single point of failure for censorship — not for funds, since force-inclusion via L1 is available)
• Open data availability on Ethereum (state can be reconstructed independently)

Stage 1 chains additionally provide:

• Permissionless or semi-permissionless proof challenges — fraud or validity proofs anyone can submit
• Trustless exit: users can withdraw to L1 without operator cooperation
• Minimum 7-day upgrade notice window

What Stage 0 means in practice: A multisig controlled by the rollup team can pause contracts or push upgrades without a time-lock. Users must trust that the team does not act adversarially. This is not a theoretical risk — it is the current operational model for zkSync Era, Linea, and Blast.

No major rollup has reached Stage 2 as of May 2026. Stage 2 requires fully permissionless proofs with no security council override outside narrow on-chain-defined recovery scenarios. Leading candidates expect to target Stage 2 in late 2026 to 2027.

Risk summary

Sequencer centralization, upgrade-key risk, Stage 0 operator trust, proof-system bugs, and tail-rollup wind-down are the main 2026 L2 risks. Last verified: 2026-05-27.

• Sequencer centralization. Every major L2 runs a single sequencer operated by the rollup team or its parent (Coinbase for Base, Offchain Labs for Arbitrum, Matter Labs for zkSync, ConsenSys for Linea). A sequencer can censor transactions or pause the chain. It cannot steal funds — force-inclusion via L1 is the recourse — but it degrades UX in an outage. Base had a 30-minute halt in September 2023; Arbitrum had a multi-hour issue in December 2023. Both resolved without loss.
• Upgrade-key risk (Stage 0 specific). On Stage 0 rollups, L1 bridge contracts are upgradeable by a multisig without a time-lock. A compromised or coerced multisig could push a malicious upgrade before users can exit. Mitigated on Stage 1 chains by the 7-day notice window.
• Proof-system risk. ZK rollups depend on the soundness of their cryptographic proof system. A circuit bug could allow a false state transition to be accepted on L1. Responsible-disclosure incidents have occurred at multiple ZK teams (including zkSync Era's proof pause); none has been exploited in production. Optimistic rollups depend on fraud-proof windows being long enough for honest challengers to act.
• Base stack migration risk. Base's departure from the OP Stack to a proprietary unified stack introduces a new risk surface: a smaller contributor base for the core rollup code, and reduced external audit coverage vs. the widely-reviewed OP Stack. The migration is in progress through 2026; edge cases during hard fork transitions deserve monitoring.
• Tail-rollup risk. 50+ of the smaller L2s launched 2023–2025 with token incentive programs are unlikely to exist as standalone chains by 2027. Token-incentive TVL is sticky for the duration of the program and evaporates afterward. If you hold assets on a tail rollup, plan an exit path before the team does.
• Bridging risk — see Best Crypto Bridges 2026. Most realized L2 losses in 2024–2026 were bridge phishing or approval exploits, not rollup protocol exploits.

Looking ahead to 2027

For the bigger picture, see where we think the next 10 years of crypto are headed. A few specific things to watch:

• Superchain interop without Base. The Superchain's shared sequencer and atomic cross-chain message bus is targeting mainnet launch later in 2026. Without Base, the addressable TVL pool for Superchain interop is smaller — but 33 remaining chains processing 17M+ daily transactions is still a significant surface. If cross-chain atomicity ships cleanly, the narrative around OP-the-token strengthens.
• Stage 2 and decentralized sequencers. Espresso, Astria, and Radius are the leading shared decentralized sequencing infrastructure projects. The first major L2 to credibly decentralize its sequencer and reach Stage 2 earns a real institutional unlock. Arbitrum and OP Mainnet are the leading candidates.
• ZK proof cost compression. Fusaka's PeerDAS already cut DA costs 40–60%. As recursion and folding schemes (Airbender, Stwo, SP1) mature, proof generation costs drop further. At near-zero proof costs, optimistic rollups lose their remaining cost advantage and the 7-day withdrawal becomes a pure liability with no upside. Most new L2 launches in 2026 are ZK.
• L2 consolidation. Post-Fusaka, several smaller rollups face near-zero sequencer margins. If blob demand rises and blob base fees harden, some tail rollups face raise-fees-or-close decisions. Expect named consolidations in 2026–2027.

Verdict

Default DeFi: Arbitrum (Stage 1, ~$19B TVL). Default consumer/onboarding: Base (12.89M daily txns). ZK with highest trust minimization: Scroll (Stage 1). ZK with AA: zkSync Era (Stage 0). Yield-on-balance: Blast (Stage 0, eyes open). Most active users hold wallets on 3–4 L2s. Last verified: 2026-05-27.

• Default L2 for DeFi: Arbitrum (Stage 1, deepest markets, BoLD permissionless fraud proofs).
• Default L2 for consumer apps and onboarding: Base (Coinbase Smart Wallet, 12.89M daily txns, passkey login).
• Default for ZK tech with Stage 1 trust: Scroll (bytecode EVM equivalence, validity proofs, no 7-day exit window).
• Default for ZK + native account abstraction: zkSync Era (Airbender prover, Stage 0 — accept the trust trade-off consciously).
• Default for yield-on-balance: Blast (native ~4% ETH / ~5% stables, Stage 0 with no fraud proofs).

Most active users hold wallets on 3–4 L2s. With cheap bridging via Across/CCTP and near-zero fees post-Fusaka, moving between them is essentially free.

---

Related: Best Crypto Bridges 2026