The code for the next version of Aztec is finished and now awaits the vote of its token holders. Proposal AZUP-3, which replaces the current rollup with a new one, reached the governance forum on October 9, 2026, and a day later the addresses of the contract bundle were published. The headline change is measurable: messages traveling from Ethereum to Aztec currently take between 24 and 108 seconds and will resolve in 12 or 30 seconds, according to Aztec itself.
Aztec is a privacy-focused layer 2 network on Ethereum. Its contracts execute private logic on the user's device and send a cryptographic proof to the sequencers, which validate the result without ever seeing the data. The network opened its Alpha phase on March 31, 2026, and activated version 5 on July 21 of the same year, when it cut the proving time of a private transaction by more than half. Now it is preparing a different kind of leap: v6 is not a revision of the rollup, but a new rollup.
The vote to activate it begins this Saturday, October 17, 2026, and lasts a week, closing on the 24th. But this is not a simple change. The new bundle shifts the reward split between sequencers and provers, opens a fee margin that governance can raise later, and sets aside 1.8 million AZTEC tokens so the previous version can keep operating for a month. Each of those pieces travels inside the twelve improvement proposals that make up AZUP-3.
A new rollup to shorten the journey from Ethereum
Today, a message leaving Ethereum for Aztec waits for the Inbox tree to close, the layer-one contract that accumulates pending sends. That journey takes between 24 and 108 seconds.
Proposal AZIP-22, dubbed Fast Inbox, has nodes incorporate each message into the block as soon as they see it on Ethereum, without waiting for the tree to close or for the two-checkpoint delay.
The result drops to a range of 12 to 30 seconds, again according to the proposal. The benefit extends to everything that crosses the border between the two networks: deposits, bridges and portal flows.
For a network that prides itself on privacy within the Ethereum ecosystem, shortening that wait also reduces the window in which the user has already sent their money and still cannot operate with it on layer two.
The improvement is real, though it should be measured precisely. Going from 24-108 seconds to 12-30 seconds does not make the bridge instant, because it eliminates the long wait and leaves the delay on the order of a block. On a network that settles its proofs on Ethereum, that is still a concrete, verifiable gain, not a slogan.
The numbers that change with v6
The bundle does not only touch the messenger. The network distributes 500 AZTEC per checkpoint, the unit with which it groups blocks and pays rewards, and v6 shifts the share each actor keeps. These are the figures published in the proposal and that any operator should check before voting.
| Change |
Before |
After |
Proposal |
| Messages from Ethereum to Aztec |
24 to 108 seconds |
12 to 30 seconds |
AZIP-22 |
| Sequencer share |
350 AZTEC (70%) |
450 AZTEC (90%) |
AZIP-30 |
| Prover pot per checkpoint |
150 AZTEC |
50 AZTEC |
AZIP-30 |
| v5 reward after the change |
500 AZTEC |
50 AZTEC for 30 days |
AZIP-34 |
| Key registry gas cap |
250.000 |
300.000 |
AZIP-33 |
| Layer 1 gas per checkpoint |
300.000 |
500.000 |
AZIP-35 |
| Layer 1 gas per epoch proof |
3.600.000 |
4.000.000 |
AZIP-35 |
| Protocol fee margin |
did not exist |
starts at zero |
AZIP-23 |
Total emissions do not change: they remain 500 AZTEC per checkpoint. What moves is the split. The sequencer's share rises from 70% to 90%, while the pot that pays the prover per block falls from 150 to 50 AZTEC per checkpoint.
What does each network actor gain and give up?
For a sequencer, the increase is direct. It goes from 350 to 450 AZTEC per checkpoint if it follows the latest rollup, and its participation moves to v6 at the moment of execution.
Anyone who prefers to stay on v5 earns 35 AZTEC per checkpoint until the previous version's reserve runs out. Their first duties on the new network arrive two or three epochs after the change.
For a prover, the adjustment goes the other way. The per-block reward pot falls from 150 to 50 AZTEC, though the revenue that comes from fees is untouched. Activity scores only rise with full epoch proofs, a change that orders the work of those competing to prove the state of the network.
AZTEC holders see no new emissions, because the total stays intact. What they gain is control. Governance receives a margin on the fee and a rewards calculator that can be changed without deploying a rollup, two levers that do not exist today and that will be activated in later proposals.
For developers, the important warning is that v6 is a different rollup. Contracts must be recompiled, class identifiers and addresses change, and layer-two fees rise slightly because of the new gas constants. None of that is resolved by a silent upgrade.
The fee margin and the rewards calculator
Until now fees were calculated at the exact cost of operating the network, with no buffer. Proposal AZIP-23 adds a margin that governance sets above that cost, with a recipient also chosen by vote.
The margin starts at zero and can only rise in limited steps approved by a later governance proposal. With that design, no one's bill changes on the first day of deployment.
The rewards calculator in AZIP-31 solves another structural problem. Previously, changing the formula that decides what a sequencer earns required deploying an entirely new rollup, with everything that entails.
Now governance can point the rollup to a separate contract. If that calculator fails or misbehaves, the system reverts to the default reward and a proof never gets stuck.
Both pieces are infrastructure for better governance, not visible improvements for the end user. Their value will appear when governance uses them, and that use is not yet written into AZUP-3. They remain as tools loaded for the future, which is exactly what they claim to be.
The twelve changes that travel inside AZUP-3
The bundle groups twelve improvement proposals, known by their AZIP acronyms. Some are big and others are low-profile technical adjustments. This is the full list published on the Aztec forum:
- AZIP-22, Fast Inbox
- AZIP-23, protocol fee margin
- AZIP-24, first prover attribution per checkpoint
- AZIP-25, activity score only for full epoch proofs
- AZIP-26, transaction effects tree in block headers
- AZIP-27, rate-limited exits for staking providers
- AZIP-29, protocol nullifier refinement
- AZIP-30, v6 deployment with a 90% sequencer share
- AZIP-31, pluggable rewards calculator
- AZIP-33, raising the GSE gas cap to 300,000
- AZIP-34, sustaining rollup v5 after the change
- AZIP-35, Glamsterdam gas constants
Two of them deserve separate attention. AZIP-26 makes each block header commit to a tree with the effects of its transactions, so that a contract can prove against that header.
AZIP-27 lets a staking provider initiate the exit of the positions it operates, something that previously only the delegator who had entrusted their tokens to it could do.
That exit carries a shared limit of 5% of the validator set every seven days. The cap exists so that no one uses withdrawals as a tool to tilt a governance vote, a risk that had already appeared in other delegated staking designs.
The real game is in the migration
The easy headline is speed. The reality is more nuanced: messages already arrived in seconds and will arrive in fewer seconds, without ceasing to depend on Ethereum's cadence. The gain is felt above all in bridges and deposits, where every second of waiting is money in transit.
What will really decide v6's success is the migration. A new rollup forces everyone operating the network to move at once, from those sequencing to those proving.
Aztec has designed its Alpha phase as a succession of deployments that do not carry state over. As already happened with the privacy experiments tested on this network, the visible part is the feature and what really costs is the infrastructure that sustains it.
If the vote passes, execution will follow in the coming days and with it the first month of coexistence between the two versions. If it does not pass, the code will remain ready and the payload will remain deployed, so the governance calendar will start again with signaling.