Ethereum’s finality reduction draft targets a fourfold cut in settlement waits

Ethereum researchers are targeting an at least fourfold reduction in finality time with a draft that would split validators’ voting duties. Filed on October 21, 2026, the Ethereum finality reduction proposal, EIP-8453, would change how the network selects its chain head and locks blocks into its history—but its speed target still depends on further parameter selection and analysis.

Key takeaways

• EIP-8453 would separate block-selection votes from finality votes.
• A roughly 512-validator committee would handle head votes.
• Activation would require an Ethereum consensus-layer hard fork.
• The proposal remains an early draft awaiting another editorial review.

Ethereum’s finality reduction proposal targets settlement waits

EIP-8453, “Decoupled Consensus,” aims to shorten the wait before an Ethereum block becomes final. Finality marks the point beyond which a block cannot be reorganized, giving users certainty that a deposit, bridge transfer or large payment has settled.

Today, Ethereum ties epoch-based finality votes to slot-by-slot fork choice—the rule used to select the current head of the chain. The voting process serves two purposes: keeping the chain moving and committing its older history.

Ben Edgington opened pull request #12566 in the ethereum/EIPs repository. The proposal’s other authors are Francesco D’Amato, Roberto Saltini, Yann Vonlanthen and Mikhail Kalinin.

Separate votes for the chain head and finality

The draft would assign head votes to a smaller committee while retaining the entire validator set for finality votes. This separates block selection from the process that makes blocks final.

For each slot, an available-chain committee of about 512 validators would vote on the chain head. A separate round of attestations from all validators would establish finality with a two-thirds quorum.

Under that design, finality votes could circulate and be combined on a separately configured schedule, rather than remaining tied to the pace of fork choice. The authors describe the purpose as separating the votes “to improve reorg resilience and enable faster finality.”

The draft also identifies better fork-choice security and groundwork for a later post-quantum transition as goals, alongside shorter finality times.

A faster settlement target, not a confirmed timetable

The Ethereum finality reduction target is conditional, not a demonstrated result. The draft makes its fourfold ambition subject to the selection and analysis of technical parameters; it does not establish a confirmed settlement time in minutes.

The authors say earlier finality would shorten the period during which blocks remain exposed to reorganization and improve user experience. For services that wait for finality before treating funds as settled, a shorter interval would reduce that wait.

Ethereum Foundation research treats funds tied up during settlement as a cost for transfer services. Faster settlement would let those services reuse their funds sooner, but the draft provides no estimate of capital freed or confirmed effect on transfer fees.

Existing transfer services already distinguish between speed and cost. Circle, the issuer of USDC, charges 0.01%, or $1 per $10,000, for its fast route from Ethereum before network and application charges. Its standard route has no Circle transfer fee. Circle also limits the value of fast transfers awaiting final settlement, with completed settlements releasing capacity for further transfers.

Implementing EIP-8453 would require a consensus-layer hard fork. Conflicting attestations would be subject to slashable penalties, meaning validators could be penalized for submitting them.

An early draft with another review ahead

EIP-8453 remains at an early stage rather than being ready for activation. Edgington wrote that “many aspects remain TBD” and that he was “primarily just looking for an EIP number assignment at this point.”

An EIP editor has since assigned number 8453. The proposal file, EIPS/eip-8453.md, still requires one additional editor review.

Article produced with the assistance of artificial intelligence and reviewed by the editorial team.