zkSync vs Starknet: EVM Compatibility, Fees, and TVL Compared

zkSync and Starknet both scale Ethereum with zero-knowledge proofs, but they solve compatibility in opposite ways. zkSync compiles standard Solidity and Vyper contracts into its own EraVM bytecode, so most Ethereum contracts deploy on zkSync with little rework. Starknet runs on Cairo, a language StarkWare built for provable computation, and Cairo was not designed around Solidity, so most Ethereum contracts need a full rewrite before they run on Starknet. That one design choice decides most of what follows: which contracts port over, how each network proves a block, and who each one attracts. Both networks already ran their token airdrops. STRK went out in February 2024, and ZK followed in June 2024. The decision today comes down to fees, throughput, ecosystem size, and 2026 roadmap plans.

EVM compatibility: the core difference

zkSync lets Solidity and Vyper developers deploy with almost no code changes, while Starknet requires learning Cairo, a language built for zero-knowledge proving. zkSync's compiler, called zksolc, compiles Solidity or Vyper through LLVM into EraVM bytecode. Researchers call this a Type 4 zkEVM. It matches Ethereum at the source-code level, but it does not match Ethereum at the bytecode level. This distinction matters more than marketing pages suggest. A contract audited on Ethereum gets recompiled into different bytecode on zkSync, and the source code stays almost the same even though the compiled output changes. Security teams still recommend a fresh audit after that move. Starknet takes the opposite path. Cairo is StarkWare's own language, and every Starknet contract is written or ported into it from scratch. That full rewrite is real friction for teams with existing Solidity code, but in exchange, Cairo was built from the ground up to be cheap to prove with STARKs. It also gives Starknet native account abstraction. Every Starknet account has been a smart contract account since genesis, with no separate externally owned account type to work around. Starknet has not ignored Solidity developers. Kakarot, a Cairo-based zkEVM built on Starknet, reached its public mainnet phase in 2025, letting Solidity contracts reach Starknet through an added layer. Tools like Hardhat, Foundry, and MetaMask can then target Starknet too, though that layer still sits between your contract and Starknet's native execution path.

Fees, proving tech, and throughput compared

zkSync currently advertises the lower per-transaction cost of the two networks. zkSync's own site quotes transfers around $0.0001, and its Airbender prover generates that proof in under a second on ordinary GPU hardware. Starknet's average transaction fee runs roughly $0.002 to $0.01, and StarkWare has an 80% fee cut planned for its 2026 roadmap. Both networks rebuilt their proving systems in 2025 and 2026. zkSync's Airbender prover and Starknet's Stwo prover both replaced older, slower systems, and StarkWare puts Stwo's throughput gain near 100 times its predecessor, a prover called Stone. On throughput, zkSync's 2026 Atlas upgrade targets sequencing above 15,000 transactions per second, along with one-second finality across its network of chains. Starknet holds a real sustained record near 127 transactions per second over 24 hours, with a burst test that once hit 857 transactions per second. Its 2026 Rust Committer upgrade targets a further 2 to 3 times gain. Both numbers describe ceilings, and neither matches everyday load. Ethereum's EVM opcodes were built to run on a virtual machine. Cheap zero-knowledge proving was never part of that design, so a prover must do extra work encoding those opcodes into arithmetic circuits. Cairo was built the other way, as a proving-first instruction set. That design gave Starknet room to add a change that touches consensus itself: validators can now secure the network with both STRK and bridged Bitcoin through strkBTC, live since May 2026. Reworking block validation to accept a second asset is a bigger structural change than adding a token to a dApp, and an EVM-equivalent chain has less room to make that kind of change.

Ecosystem, TVL, and token status compared

Starknet currently holds more total value locked than zkSync Era on most trackers. Independent trackers put Starknet's TVL between roughly $300 million and $600 million in mid-2026, a figure that roughly doubled off a November low, and Starknet has climbed back toward the top ten Layer 2 networks by that measure. zkSync Era's TVL sits lower on the same trackers, in a range of roughly $100 million to $200 million, well down from an earlier peak near $500 million. zksync.io itself cites a combined figure near $4 billion, but that figure spans its whole Elastic Network of 18-plus chains. It is not a total for the Era chain alone. zksync.io also reports more than 700 million cumulative transactions across that same network. The two native tokens launched close together in 2024 but serve different roles on their own chains. STRK went live first, with roughly 700 million tokens airdropped to about 1.3 million wallets starting February 20, 2024. ZK followed on June 17, 2024, with a 3.675 billion token airdrop that ranks among the largest Layer 2 airdrops on record. Both tokens fund gas and governance today, and STRK also secures the network directly through Starknet's staking system. ZK Nation runs zkSync's governance through a staking pilot launched in February 2026, aimed at laying groundwork for a future decentralized sequencer, though neither sequencer is fully decentralized yet. Starknet shipped a decentralized sequencer architecture in its v0.14 release, and it already has staking live, with further phases planned through 2026 to decentralize block validation too. zkSync's sequencer is still run by Matter Labs, and its 2026 staking pilot is framed only as a first step; no handoff to a decentralized sequencer has happened yet.

Which network to build on or hold assets on

Choose zkSync if your team already has a working Solidity codebase, since it gives you the smallest possible rewrite of the two networks. Enterprise projects also lean toward zkSync's Prividium offering, which lets an organization run a private, permissioned chain that still settles to Ethereum. Skip zkSync if you need a fully decentralized sequencer right now, or if your project needs account abstraction without an EVM adapter layer. Choose Starknet if you are building something Cairo does well, including high-frequency trading, gaming logic, or native account abstraction. Cairo also gives you STARK-based proving without carrying EVM baggage, and Starknet suits teams open to using Kakarot for a partial EVM path while keeping the option open to write Cairo-native contracts later. Skip Starknet if your team cannot spend time learning Cairo, or if your contracts depend on a Solidity library with no Cairo equivalent yet. For a wallet holder deciding where to bridge assets, both token airdrops are long over, so speculation about a future drop is not a good reason to move funds today. Compare current fees on the specific action you plan to take, and check which dApps you actually want to use on each chain. Starknet currently carries more total value locked, and some users read that as a rough signal of battle-tested liquidity. A regulation change could shift that picture, and so could a Kakarot mainnet milestone or either network finishing sequencer decentralization first. That is the moment to revisit this comparison, since today's numbers will not hold forever.

Frequently asked questions

Sources

• ↗zkSync official site
• ↗Starknet official site
• ↗ZKsync fee structure documentation
• ↗Starknet decentralization roadmap
• ↗ZK Nation token documentation

Related guides