FinWiki
日EN中

ZK-EVM Rollup Maturity Comparison Matrix 2026 · proving system / TVL / DA / decentralization across 9 rollups

Confidence: Likely Updated 2026-05-26 Review by 2026-11-25 Sources 12 Machine-translated Original (JA)
#systems#matrix#zk-evm#rollup#ethereum#l2
On this page

TL;DR

  • A comparison of 9 active ZK-EVM / ZK-rollups across the 9 axes of EVM equivalence · proving system · proof time · gas cost · TVL · mainnet maturity · decentralization · ecosystem · DA layer as of 2026-Q2
  • EVM equivalence ladder (Vitalik Type 1-4): Linea/Scroll → Type 2(EVM-equivalent · 99%) · Polygon zkEVM → Type 2.5 → evolving toward Type 1 · zkSync Era → Type 4(EVM-compatible · different bytecode) · StarkNet → Type 4(Cairo native · Solidity via Kakarot) · Taiko → Type 1(fully EVM-equivalent · base-rollup model) · Miden → non-EVM (private execution + Solidity bridge)
  • Proving system 5 major mainstreams: Plonky2(Polygon · STARK→SNARK recursion) · Boojum (zkSync · STARK + FRI) · Cairo-STARK (StarkNet) · Halo2 / Recursive Halo2(Scroll · Taiko · PSE lineage) · EVM Prover (Linea · zk-SNARK with vk2.0)
  • TVL 2026-Q2 estimates: zkSync Era ~$2.0B · Linea ~$1.2B · Polygon zkEVM ~$800M · Scroll ~$700M · StarkNet ~$650M · Taiko ~$350M · Immutable zkEVM ~$280M · Miden ~$30M · Kakarot ~$15M (early)
  • DA layer choice: the majority still use L1 blobs (EIP-4844 · see pectra-eip-7691-blob-l2-impact) · a few choose alt-DA: Polygon zkEVM offers a validium mode · StarkNet Volition · Miden’s own P2P · Kakarot exploring Celestia · Linea/Scroll/Taiko default to L1
  • routing: compared with systems index · cross-chain-five-pole-comparison-matrix (inter-rollup messaging selection)

Wiki route

This entry sits under systems index. Read it against Pectra EIP-7691 · blob doubling and the L2 economic chain for the L1 blob-capacity context that drives rollup gas cost, and Pectra upgrade overview for the broader Ethereum upgrade landscape. Pair with Vitalik L1/L2 strategy anchor for the meta-strategy that frames “L2 carrying 99% of users”. For proving system depth, see multi-proof aggregation fault prover and formal spec implementation codesign. For long-term cryptographic risk, see post-quantum blockchain day-1 integration. For restaking-backed DA (EigenDA serves several rollups in this matrix), see EigenLayer overview. Cross-link to cross-chain five-pole comparison matrix when reasoning about L2-to-L2 messaging.

Why this matrix matters

After ZK-EVM rollups went live on mainnet intensively in 2023-2025 年· they evolved from a “technical-route war” into a multi-objective optimization space of “EVM equivalence vs proving cost vs decentralization vs ecosystem traction”. When developers select which ZK rollup to deploy to · facing 6-8 candidates (each claiming “most EVM-equivalent · fastest · cheapest”) · without a unified comparison table they will be misled by market marketing.

Institutional customers (see the J.P. Morgan Kinexys / DTCC / SWIFT pilot in CCIP institutional) likewise need to decide “on which L2 to settle tokenized assets”. What institutions care about is not TPS but · the decentralization of the prover/sequencer (currently the sequencer of every ZK-EVM is centralized · and provers are mostly centralized) · upgrade governance (being able to upgrade the smart contract within the challenge window = actually controlling the funds) · and DA layer risk (when using validium · data availability depends on an external committee).

This matrix compares 9 publicly-running ZK-EVM / ZK-rollups · spread across 9 axes · with a 2026-Q2 data snapshot. Optimistic rollups (Arbitrum / Optimism / Base / Blast) are not included in this matrix · see a separate optimistic matrix (its publication timing is not specified in the public materials cited).

Per-rollup sections

Polygon zkEVM / AggLayer

EVM equivalence: at the 2023 年launch it was Type 3(EVM-compatible · some opcode differences) · gradually evolving to Type 2.5 in 2024-2025 年. The Polygon team’s roadmap is Type 1 full equivalence.

Proving system: Plonky2(STARK recursion → SNARK final · GPU-friendly). Upgraded to Plonky3 in 2024 年for ~30% speedup.

Proof time: single batch ~5-10 minutes (2026-Q2). L1 submit average ~30 minutes finality.

Gas cost: user sub-cent ($0.001-0.01 simple transfer).

TVL May 2026: ~$800M (estimate · including the AggLayer federation).

Mainnet maturity: 2023.03 mainnet beta · 2024 Stage 1 decentralization · 2026 is also Stage 1(Sequencer centralized · Prover partially opened).

Decentralization: Sequencer: operated centrally by Polygon Labs · plans to introduce the Espresso shared sequencer in 2026-2027 . Prover: 5 社’s prover network (Polygon Labs + 4 third parties). Upgrade: 7-day timelock + Security Council 7-of-13.

Ecosystem: ~150 dApp · including QuickSwap · Polygon ID · Astar zkEVM · multiple enterprise cases (Franklin Templeton tokenized fund pilot).

DA layer: default L1 blobs (EIP-4844). Offers a validium mode (DAC 7-of-9 multisig · for enterprise private chains). AggLayer is Polygon’s cross-chain unifier · multiple zkEVMs can share settlement.

zkSync Era (Matter Labs)

EVM equivalence: Type 4(EVM-compatible · different bytecode · Solidity via its own zkSolc compiler). Specialized dev tooling is required.

Proving system: Boojum (STARK + FRI · GPU-friendly). Upgraded to Boojum v1.5 in 2024 年for ~40% efficiency gain. Final SNARK wrap on Ethereum.

Proof time: ~10-15 minutes batch · L1 submit ~30-45 minutes finality.

Gas cost: user sub-cent ($0.001-0.005). Account Abstraction native (no ERC-4337 needed).

TVL May 2026: ~$2.0B (estimate · including ZK Stack hyperchains).

Mainnet maturity: 2023.03 mainnet · announced the ZK Stack framework in 2024 · 2026 Stage 1 decentralization.

Decentralization: Sequencer: Matter Labs centralized · plans to introduce a decentralized sequencer in 2026 Q3-Q4 (Espresso partnership). Prover: Matter Labs + 1-2 third parties. Upgrade: 21-day timelock (extended from 1-day after 2024 ) + Security Council 9-of-15.

Ecosystem: ~250 dApp · SyncSwap · Maverick · Aave v3 · Argent native AA · ZK Stack hyperchains (GRVT · Treasure · Cronos zkEVM).

DA layer: default L1 blobs. The validium mode (“zkPorter”) is planned but deployment is deferred. ZK Stack hyperchains can choose EigenDA / Celestia.

StarkNet

EVM equivalence: Type 4(Cairo native · not Solidity). Solidity compatibility is via the Kakarot subproject (EVM-on-StarkNet · not fully production grade) or the Warp transpiler (deprecated). Cairo is a Turing-complete language optimized for STARK.

Proving system: Cairo-STARK (native STARK · no SNARK wrap needed · but L1 verification cost is on the higher side). STARK is quantum-resistant (see post-quantum blockchain day-1).

Proof time: ~30-60 minutes batch (STARK proof size is large and slower).

Gas cost: user ~$0.01-0.1(higher than zkSync/Polygon · due to STARK proof size). The 2025 stwo prover upgrade targets -50% cost.

TVL May 2026: ~$650M (estimate).

Mainnet maturity: 2021.11 alpha · 2023 v0.13 mainnet · 2024 v0.14 Volition introduced · 2025 Stage 1 partial.

Decentralization: Sequencer: StarkWare centralized · plans to decentralize via a PoS validator set in 2026-2027 . Prover: StarkWare-led · third-party prover SDK opened. Upgrade: Foundation governance + STRK token voting.

Ecosystem: ~120 dApp · including Ekubo · JediSwap · ZkLend · zkLend · Briq · Realms. Argent / Braavos AA wallet native support. Institutional cases: Visa Aeon · Provenance pilot.

DA layer: L1 blobs (default) + Volition mode (user chooses on-chain or DAC off-chain) + EigenDA under exploration.

Scroll

EVM equivalence: Type 2(EVM-equivalent · 99% bytecode-level compatible · almost all Solidity dApps can deploy with zero modification).

Proving system: Halo2 + PSE-flavor zkEVM (based on the long-term research of Privacy Scaling Explorations). Upgraded to multi-proof aggregation in 2025 年(see multi-proof aggregation).

Proof time: ~15-25 minutes batch · L1 submit ~30 minutes finality.

Gas cost: user sub-cent ($0.005-0.02).

TVL May 2026: ~$700M.

Mainnet maturity: 2023.10 mainnet · 2024-2025 continuously audited · 2026 Stage 1.

Decentralization: Sequencer: Scroll team centralized · plans to decentralize in 2026-2027 . Prover: opened to third parties (Scroll prover network beta). Upgrade: 7-day timelock + Security Council.

Ecosystem: ~180 dApp · Aave · Uniswap · Pendle · GMX and other mainstream all deployed. Scroll Canvas is a user-onboarding tool.

DA layer: default L1 blobs. The Scroll team has expressed a clear stance of not doing validium · maintaining L1 DA purity (EVM-equivalent + L1 DA is an institutional selling point).

Linea (Consensys)

EVM equivalence: Type 2(EVM-equivalent · 99% compatible, similar to Scroll). Consensys emphasizes “drop-in MetaMask / Truffle / Infura toolchain compatibility”.

Proving system: zk-SNARK custom prover (Linea VK2.0 · based on Vortex + Arcane). vk3.0 upgrade in 2025 年improves proof speed. Recursive proof aggregation.

Proof time: ~10-20 minutes batch · L1 submit ~30 minutes finality.

Gas cost: user sub-cent ($0.002-0.01).

TVL May 2026: ~$1.2B (estimate · boosted by MetaMask traffic).

Mainnet maturity: 2023.07 mainnet · 2024 v1.5 upgrade · 2026 Stage 1.

Decentralization: Sequencer: Consensys centralized · plans to decentralize in 2026-2027 . Prover: Consensys-led · third-party prover SDK opened. Upgrade: 7-day timelock + Security Council.

Ecosystem: ~200 dApp · Aave · Uniswap · LayerZero · Stargate · OKX · the deep integration with MetaMask is the largest traffic entry point.

DA layer: default L1 blobs. Consensys continues an “all on-chain” stance · does not offer validium.

Taiko

EVM equivalence: Type 1(fully EVM-equivalent · bit-for-bit · including EVM opcode / precompile / consensus rules). Any Ethereum dApp / L1 client can be forked directly.

Proving system: Halo2(SGX + ZK multi-prover · introduced a 3 -way prover competition of SP1 · Risc Zero · Aligned in 2024 年). Pioneer of multi-proof aggregation (see multi-proof-aggregation-fault-prover).

Proof time: ~20-40 minutes batch (proving is slow at 100% EVM precision because of Type 1 ).

Gas cost: user ~$0.01-0.05(slightly higher than Scroll/Linea · due to the prove cost of Type 1 ).

TVL May 2026: ~$350M (estimate).

Mainnet maturity: 2024.05 Alpha-6 mainnet · 2024.06 Hekla testnet · 2025 mainnet stable · 2026 Stage 1 evolving.

Decentralization: Sequencer: Based rollup (borrows the L1 proposer as the sequencer · the only fully decentralized sequencer among ZK-EVMs). Prover: opened as a permissionless prover market (anyone can earn TAIKO by submitting a proof). Upgrade: 7-day timelock.

Ecosystem: ~80 dApp · TaikoSwap · mainly Ethereum-native dApp forks. The based-rollup concept is the academic core of the ETH community · publicly recommended by Vitalik.

DA layer: default L1 blobs (a based rollup requires L1 DA).

Kakarot

EVM equivalence: Type 4(EVM implemented in Cairo · running on top of StarkNet). Theoretically 99% Solidity-compatible · but with edge-case differences.

Proving system: borrows StarkNet’s Cairo-STARK (the Kakarot smart-contract layer is Cairo · internally interpreting EVM bytecode).

Proof time: depends on StarkNet (~30-60 minutes).

Gas cost: depends on StarkNet · slightly higher than native StarkNet.

TVL May 2026: ~$15M (early adoption).

Mainnet maturity: 2024-Q4 mainnet beta · 2025-2026 trial operation · not production grade.

Decentralization: borrows StarkNet governance · the Kakarot team leads smart-contract upgrades.

Ecosystem: ~10-20 dApp · mainly developer experimental deploys. Kakarot’s strategic value is “bringing the EVM into StarkNet” · it is not an independent L2 .

DA layer: borrows StarkNet (L1 blobs + Volition + EigenDA under exploration).

Polygon Miden

EVM equivalence: non-EVM · Miden Assembly (MASM) · a STARK-friendly ISA. The EVM is via the Miden-EVM transpiler (early). The design philosophy is “client prover · the user proves the transaction locally · then submits the proof to the sequencer” (in contrast to the conventional sequencer-prove model).

Proving system: Miden VM (based on the Winterfell STARK framework). Client-prover-friendly (proving possible on a laptop in seconds to minutes · depending on transaction complexity).

Proof time: user client ~1-30 seconds (simple transaction) · L1 submit batch ~10-20 minutes.

Gas cost: user sub-cent · because proving is local · L1 pays only for verification + DA.

TVL May 2026: ~$30M (early).

Mainnet maturity: 2024 testnet · 2026 mainnet beta · not production grade for high-value DeFi.

Decentralization: Sequencer: Polygon Labs centralized (early). Prover: user self-proving is the core of the design · naturally decentralized (the user controls their own transaction proof).

Ecosystem: ~10 dApp · mainly private-payment / DeFi experiments (Miden emphasizes client-side privacy as its differentiator).

DA layer: Miden P2P own DA (experimental · not L1 blobs). Plans to explore Celestia / EigenDA as fallback. Polygon AggLayer connection.

ImmutableX (Immutable zkEVM)

EVM equivalence: Type 2(based on the Polygon zkEVM stack · EVM-equivalent 99%). A gaming-specialized ZK-rollup.

Proving system: Plonky2(borrows the Polygon zkEVM proving infrastructure).

Proof time: ~10-20 minutes (inherits Polygon zkEVM performance).

Gas cost: user gas-free for NFT mint / transfer (the Immutable team subsidizes · gaming UX priority) · other transactions sub-cent.

TVL May 2026: ~$280M (NFT-heavy · not fully fungible TVL).

Mainnet maturity: 2024-Q1 zkEVM mainnet (previously ImmutableX 1.0 was a StarkEx validity pool · not zkEVM). 2025-2026 gaming ecosystem expansion.

Decentralization: Sequencer: Immutable centralized (gaming positioning · does not prioritize decentralization). Prover: borrows the Polygon prover network. Upgrade: Immutable Foundation + 7-day timelock.

Ecosystem: specialized in web3 gaming —— Illuvium · GuildFi · Gods Unchained · Ember Sword · Counterplay and other 80+ games. The most dominant in the gaming vertical among ZK-EVMs.

DA layer: L1 blobs default · some NFT metadata via validium (DAC mode) to reduce cost.

Big comparison matrix table

9 ZK-EVM / ZK-rollups × 9 axes comparison (2026-Q2 status):

RollupEVM TypeProving SystemProof TimeGas CostTVL May 2026Mainnet MaturitySeq/Prover DecentralizationEcosystemDA Layer
Polygon zkEVM / AggLayerType 2.5(target Type 1)Plonky2/3(STARK→SNARK)5-10 minsub-cent$800M2023-03 · Stage 1Seq centralized · Prover 5 社network · 7d timelock + SC 7/13~150 dApp · QuickSwap · Polygon ID · Franklin Templeton pilotL1 blobs + Validium mode · AggLayer unifier
zkSync EraType 4(zkSolc)Boojum (STARK+FRI)10-15 minsub-cent ($0.001-0.005) · AA native$2.0B2023-03 · Stage 1Seq centralized · Prover Matter Labs+1-2 · 21d timelock + SC 9/15~250 dApp · SyncSwap · Aave · Argent · ZK Stack hyperchainsL1 blobs · Validium (“zkPorter”) deferred · hyperchains can choose EigenDA/Celestia
StarkNetType 4(Cairo native · Solidity via Kakarot)Cairo-STARK (native · quantum-resistant)30-60 min$0.01-0.1$650M2021-11 alpha · 2023 v0.13 · Stage 1 partialSeq centralized · Prover SDK opened · Foundation+STRK governance~120 dApp · Ekubo · ZkLend · Visa Aeon · Provenance pilotL1 blobs + Volition mode + EigenDA under exploration
ScrollType 2(99% bytecode-equivalent)Halo2 + PSE-zkEVM · multi-proof agg15-25 minsub-cent ($0.005-0.02)$700M2023-10 · Stage 1Seq centralized · Prover third-party beta · 7d timelock + SC~180 dApp · Aave/Uni/Pendle/GMX · Scroll CanvasL1 blobs only (explicitly no validium)
Linea (Consensys)Type 2zk-SNARK custom (VK2.0/3.0 · Vortex+Arcane)10-20 minsub-cent ($0.002-0.01)$1.2B2023-07 · Stage 1Seq centralized · Prover Consensys+SDK · 7d timelock + SC~200 dApp · MetaMask integration · Aave/Uni/OKXL1 blobs only (all on-chain stance)
TaikoType 1(bit-for-bit EVM)Halo2(SGX+ZK) · SP1/Risc Zero/Aligned 3 -way20-40 min$0.01-0.05$350M2024-05 mainnet · Stage 1 evolvingBased rollup · Seq borrows L1 proposer · Permissionless prover market · 7d timelock~80 dApp · Ethereum-native fork · Vitalik-recommended based rollupL1 blobs (based rollup requires L1 DA)
KakarotType 4(EVM on Cairo on StarkNet)borrows StarkNet Cairo-STARK~30-60 min (borrows StarkNet)slightly higher than StarkNet$15M2024 Q4 beta · not production-gradeborrows StarkNet · Kakarot team-led~10-20 dApp · experimental deployborrows StarkNet (L1 blobs + Volition + EigenDA exploration)
Polygon Midennon-EVM (MASM · EVM transpiler early)Miden VM (Winterfell STARK) · client proveruser local 1-30s · L1 batch 10-20 minsub-cent (prove local)$30M2024 testnet · 2026 mainnet betaSeq centralized · Prover user self-prove (naturally decentralized)~10 dApp · private-payment/DeFi experimentsMiden P2P own DA · Celestia/EigenDA exploration · AggLayer connection
ImmutableX zkEVMType 2(based on Polygon zkEVM stack)Plonky2(borrows Polygon)10-20 minNFT gas-free subsidy · other sub-cent$280M (NFT-heavy)2024 Q1 zkEVM mainnet · previously StarkExSeq centralized (gaming positioning) · Prover borrows Polygon · 7d timelockspecialized in gaming · 80+ games: Illuvium/Gods Unchained/GuildL1 blobs + some NFT metadata validium

How to read the matrix:

  • EVM equivalence vs proving cost tradeoff: Type 1(Taiko) is the most expensive but has perfect compatibility · Type 4(zkSync · StarkNet · Kakarot) is the cheapest but requires dev tooling adaptation
  • DA layer choice reflects an institutional vs DeFi stance: Scroll/Linea/Taiko defend L1 blobs to the death (institutional grade) · Polygon/StarkNet/Miden offer validium/Volition (DeFi UX priority) · Immutable is hybrid (NFT uses validium to reduce cost)
  • The true state of decentralization: the sequencer of every ZK-EVM is centralized (2026-Q2) · the only exception is the Taiko based rollup. Progress on prover decentralization: Polygon (5 社network) > Scroll/StarkNet (SDK opened) > the rest (team-led)
  • TVL and ecosystem traction are not fully correlated: zkSync $2.0B high TVL but Type 4 limits developer migration cost; Linea $1.2B borrows the MetaMask traffic advantage; Scroll/Polygon Type 2 + medium TVL is the “balanced choice”

Boundary cases / future trajectory

Direction of evolution of the EVM equivalence ladder: Polygon zkEVM has a public roadmap to achieve Type 1 in 2027 年. Taiko is already Type 1 · the reference benchmark for other ZK-EVMs. zkSync Era and StarkNet maintain the Type 4 philosophy (optimizing prove cost · accepting dev tooling cost). Linea and Scroll are stable at Type 2 · an upgrade to Type 1 is not planned (judging that the cost-benefit of the last 1% of compatibility is low).

Proving system convergence vs diversification:

  • Convergence trend: most ZK-EVMs ultimately SNARK-wrap (due to L1 verification cost · groth16 / plonky etc.) · the same holds even for internal STARK (zkSync Boojum · Polygon Plonky2).
  • Diversification trend: StarkNet / Miden hold firm to native STARK · for quantum resistance + client-friendliness (see post-quantum-blockchain-day1-integration). The generalization of provers such as SP1 / Risc Zero / Aligned (any rollup can choose) is a new trend in modular proving.
  • Via Multi-proof aggregation (see multi-proof aggregation fault prover) · 1 rollup can run multiple independent provers simultaneously · with fallback on error —— Taiko is the pioneer.

Sequencer decentralization timetable: Polygon zkEVM · zkSync Era · Scroll · Linea have all publicly stated they will introduce a shared sequencer (Espresso / Astria / Radius) in 2026-2027 . The Taiko based rollup has resolved it naturally. Vitalik publicly proposed in 2024 年that ZK-EVMs should prioritize solving sequencer decentralization rather than the prover · because sequencer concentration = greater liveness risk.

AggLayer / ZK Stack / Superchain faction war:

  • Polygon AggLayer federates Polygon zkEVM + zkSync (partial) + Astar + others · sharing a settlement layer + cross-chain LP
  • zkSync ZK Stack is the Matter Labs-led hyperchain framework · has already deployed GRVT · Treasure · Cronos zkEVM
  • Optimism Superchain (not included in this matrix · OP-rollup) is the counterpart
  • In the long term some kind of unifier is needed among all rollups · but AggLayer / ZK Stack / Superchain each push forward separately · for users to use across stacks a general bridge is still required (see cross-chain five-pole matrix)

The relationship between the DA layer and EigenLayer: EigenDA (see EigenLayer overview) provides alt-DA to multiple ZK Stack hyperchains · StarkNet Volition · Miden / Kakarot etc. But whether ZK-EVMs should use alt-DA (giving up L1 security) is still debated —— the Scroll / Linea / Taiko camp holds firm to L1 DA · the Polygon / StarkNet / Miden camp offers options. Institutional customers (SWIFT / DTCC · see chainlink-ccip-institutional-messaging) explicitly require L1 DA.

**Impact of Pectra EIP-7691 ** (see Pectra EIP-7691 · blob doubling and the L2 economic chain): blob target 3→6 + max 6→9 will further reduce ZK-EVM submit cost by -50% · all rollups using L1 blobs benefit directly. 2026-Q4 Fusaka PeerDAS gives a further 10× capacity.

Formal verification progress: ZK-EVM is a natural stage for formal spec implementation co-design (see formal spec implementation codesign) —— the dual-spec consistency of Solidity / EVM spec + ZK circuit spec is the core security point. The Linea and Consensys Diligence teams have published formal-verification progress. Scroll collaborates academically with PSE over the long term. The zkSync Boojum prover is independently audited by Galois.

Institutional vs DeFi selection:

  • Institutional (tokenized fund · DvP · regulatory compliance): Linea (MetaMask Institutional integration · backed by Consensys) · Scroll (EVM-equivalent + L1 DA purity) · Polygon zkEVM (the Franklin Templeton pilot precedent)
  • DeFi-native (high TPS · long-tail dApp): zkSync Era · StarkNet · Polygon AggLayer
  • Gaming-specific: ImmutableX
  • Privacy-first: Polygon Miden (client prover + privacy design)
  • Ethereum-aligned (maximum decentralization · Vitalik-recommended): Taiko based rollup

Sources