Skip to content

Guide · Rollups

Rollup types explained: optimistic vs ZK, and what really differs

Optimistic vs ZK rollups: how fraud proofs and challenge windows differ from validity proofs, what each costs, and why the label alone does not tell you how secure a given L2 is. Not financial advice.

By Etherelysium Editorial TeamUpdated 18 November 20267 min read

Rollup types explained: optimistic vs ZK, and what really differs
Share X LinkedIn

Both optimistic and ZK rollups do the same basic job: execute transactions off Ethereum’s base layer, then post a compact proof that their results are correct back to L1. What differs is how each one proves correctness, how long you wait before that proof settles the question, and what it costs to generate. Neither label tells you how decentralized or battle-tested a specific rollup is — that is a separate question, covered in the companion guide on L2 security models.

Optimistic rollups: assume correct, then let anyone object

An optimistic rollup posts its new state root to L1 and assumes it is correct by default. It does not prove correctness up front. Instead, it opens a challenge window — commonly measured in days — during which any party running a verifier can submit a fraud proof showing that a specific posted state transition was wrong. If nobody successfully challenges it within that window, the state root is treated as settled on L1. If a fraud proof succeeds, the incorrect state transition is rolled back and the party that proposed it is typically penalized.

The practical consequence is a real tradeoff between speed and certainty. Funds leaving an optimistic rollup through its canonical bridge typically cannot be withdrawn to L1 with full security until the challenge window has passed, because that window is the actual security mechanism, not a formality. Many optimistic rollups offer faster “liquidity provider” exits for users willing to accept a third party’s assumption of that risk in exchange for speed, which is a different trust model than waiting out the window yourself.

Fraud proofs only work if someone is actually watching and willing to submit one. That is why the number of parties permitted to submit a fraud proof — and whether that set is open to anyone or restricted to an allowlist — is one of the clearest signals of how mature a given optimistic rollup’s security actually is, independent of the “optimistic” label itself.

ZK rollups: prove correct before anyone can object

A ZK rollup takes the opposite approach: it generates a cryptographic validity proof — a succinct mathematical proof that the claimed state transition follows correctly from the prior state and the batch of transactions — before the new state root is accepted on L1. A smart contract on L1 checks that proof directly. If the proof verifies, the new state is accepted immediately; there is no multi-day waiting period for anyone to object, because the proof itself is the objection-proof guarantee.

This removes the days-long challenge window that optimistic designs rely on, but it shifts the cost elsewhere: generating a validity proof for a complex batch of general-purpose transactions is computationally expensive, and building a prover fast and cheap enough to keep up with real transaction volume, across arbitrary smart contract logic rather than a narrow set of operations, has been one of the harder engineering problems in the space. Different ZK rollups make different tradeoffs between how general the computation they can prove is, how long proof generation takes, and how expensive verifying that proof is on L1.

What actually differs in practice

The trust assumption while waiting

An optimistic rollup’s security during the challenge window depends on at least one honest, capable party being willing and able to submit a fraud proof in time. A ZK rollup’s security depends on the soundness of its cryptographic proof system and the correctness of the verifier contract that checks it on L1. These are genuinely different kinds of trust, not just different speeds dressed up differently.

Withdrawal latency to L1

Optimistic rollups generally impose a real wait — the challenge window — before a withdrawal is secure without relying on a third party. ZK rollups can finalize as soon as a validity proof is generated and verified, which in well-engineered systems can be meaningfully faster, though proof generation time itself is not instantaneous and varies by design.

Cost structure

Optimistic rollups mainly pay for posting data and, occasionally, for resolving a dispute. ZK rollups additionally pay the computational cost of generating proofs, plus the gas cost of verifying those proofs on L1. Which design ends up cheaper for end users depends on transaction volume, proof system efficiency, and how both designs make use of the blob fee market described in the companion guide on blob fees and data availability.

The label is not the whole security story

“Optimistic” and “ZK” describe the proof mechanism, not the rollup’s overall decentralization or the power its operator retains over upgrades, the sequencer, or an emergency pause. A ZK rollup with a fully trusted upgrade key is not automatically safer than an optimistic rollup with a well-tested, permissionless fraud proof system and a hardened security council — and vice versa. Evaluating an L2 means looking past this label at who can submit proofs, how long users can exit if something goes wrong, and what powers remain centralized, which is exactly the subject of the companion guide on L2 security models, proofs, and escape hatches.

A worked comparison

Say the same simple transfer runs on a representative optimistic rollup and a representative ZK rollup. On the optimistic rollup, the transfer is included in a batch almost immediately and the user sees it reflected in their balance right away — but the recipient relying on a trust-minimized L1 withdrawal would, in the most conservative reading, wait out the full challenge window before treating that balance as settled against L1. On the ZK rollup, the transfer is similarly reflected immediately, and once the batch containing it is proven and that proof verifies on L1 — a process measured in minutes to hours depending on the system, not days — the state is settled without anyone needing to object. Both give the user a fast UX; they differ in how long it takes for an L1-level guarantee to catch up with that UX, and in what has to go right for that guarantee to hold.

Hybrid and evolving designs

The optimistic-versus-ZK split is not always a clean binary in production. Some rollups combine elements of both, or use validity proofs for parts of their execution while retaining an optimistic-style dispute path as a fallback. Proving technology has also been moving quickly enough that a rollup’s current proof system and challenge-window length are details worth checking directly against its own documentation rather than assuming they match whatever was true when the project first launched.

None of this changes the core distinction, though: an optimistic rollup’s settled state is a claim that went unchallenged for a defined period, while a ZK rollup’s settled state is a claim that was mathematically checked before being accepted. Knowing which kind of claim you are relying on is more useful than the “optimistic” or “ZK” label by itself.

Risk note

Neither rollup type is inherently safer in every dimension. A rollup’s proof mechanism, its challenge window or proof-generation time, and the centralization of its upgrade keys and sequencer are separate questions, and a reader comparing two L2s should check all of them rather than assuming the label settles the matter.

Not financial advice. Nothing on this page is a recommendation to buy, sell, stake, bridge, or avoid any asset or protocol.

FAQ

Which is faster to withdraw from, optimistic or ZK?

ZK rollups can finalize a withdrawal as soon as a validity proof verifies on L1, without a multi-day wait. Optimistic rollups generally require their challenge window to close first for a fully trust-minimized exit, though many offer faster third-party exits for a fee.

Does ZK mean private transactions?

No. ZK here refers to the proof technique used to verify correctness succinctly, not to transaction privacy. Most ZK rollups in production today post transaction data openly; privacy is a separate feature some systems add on top.

Can a rollup change proof types later?

Some rollups have migrated their proving approach over time as the technology matured, but doing so is a significant engineering undertaking, not a simple setting to flip.

Is a ZK rollup automatically more secure than an optimistic one?

No. The proof mechanism is one input among several. Upgrade key centralization, sequencer control, and who can submit a fraud or validity challenge all matter independently of which proof type a rollup uses.

Frequently asked questions

Which is faster to withdraw from, optimistic or ZK?
ZK rollups can finalize a withdrawal as soon as a validity proof verifies on L1, without a multi-day wait. Optimistic rollups generally require their challenge window to close first for a fully trust-minimized exit, though many offer faster third-party exits for a fee.
Does ZK mean private transactions?
No. ZK here refers to the proof technique used to verify correctness succinctly, not to transaction privacy. Most ZK rollups in production today post transaction data openly; privacy is a separate feature some systems add on top.
Can a rollup change proof types later?
Some rollups have migrated their proving approach over time as the technology matured, but doing so is a significant engineering undertaking, not a simple setting to flip.
Is a ZK rollup automatically more secure than an optimistic one?
No. The proof mechanism is one input among several. Upgrade key centralization, sequencer control, and who can submit a fraud or validity challenge all matter independently of which proof type a rollup uses.

Educational content, human-edited and possibly AI-assisted. It describes how systems work and is not financial, legal or tax advice. Report an error via our corrections page.

More in Rollups

Related articles

See all