Solana’s Alpenglow is live, but 150ms remains a target

Anza’s Alpenglow consensus upgrade went live on Solana mainnet on Oct. 21, replacing TowerBFT and cutting the time for a transaction to become irreversible from seconds to a fraction of a second. Anza’s mainnet announcement describes 150 milliseconds as the new finality time; early measurements from a public node put the median at about 404 milliseconds. That is a substantial reduction, but the target and the observed result are not the same figure.

The handoff took place at slot 459,653,000, around 14:00 UTC, according to Anza. Finality is the point at which a transaction can no longer be reversed, a stronger assurance than a wallet showing that a transaction has been received or tentatively confirmed. Faster finality could let exchanges credit deposits and merchants accept payments sooner. It changes settlement confidence, rather than the speed at which users submit transactions.

Why is the 150ms figure ahead of measured results?

Because 150 milliseconds is Anza’s target, while outside measurements include the time a public node takes to receive and process a block. OpenChainBench’s 24-hour measurement through Oct. 23 found a 404ms median, a 579ms 90th percentile and a 765ms 99th percentile across more than 357,000 slots. Its pre-upgrade median was about seven seconds under TowerBFT.

That before-and-after comparison is narrower than the headline comparison of 12.8 seconds to 150 milliseconds. The 12.8-second figure assumes 400-millisecond slots; Solana’s slots had already shortened before Alpenglow activated. Against the measured seven-second baseline, 404 milliseconds is about a 17-fold improvement. OpenChainBench’s readings reflect what a client sees from a public RPC node, not necessarily how quickly validators inside the voting exchange reach finality.

How does Votor change Solana’s settlement process?

Votor, Alpenglow’s live voting component, replaces on-chain vote transactions with direct validator messages that are combined into certificates. Under TowerBFT, votes accumulated across 32 confirmed slots before a block became final. With Votor, a block can finalize in one voting round if 80% of stake votes for it, or in two rounds when certificates reach 60%.

Moving votes out of blocks frees space that those transactions used, but it also changes what network activity counts look like. OpenChainBench estimates vote transactions made up about 62% of processed transactions in the week before the switch, so raw transaction-per-second figures from before and after Oct. 21 are not directly comparable. A drop in that headline count alone would not show that user activity had fallen.

What should users and validators watch next?

The trade-off is faster settlement with a protocol and operating change beneath it. Alpenglow replaces Solana’s consensus layer while leaving transaction execution, programs and fees unchanged, according to OpenChainBench. Validators must run compatible software and fund a 0.8 SOL admission ticket each epoch; those requirements add operational costs even as off-chain voting reduces the need to put votes in blocks.

For users, the practical test is whether public-node finality stays below a second under ordinary network conditions and whether applications begin relying on the stronger, faster finality state. For the protocol, the next signal is Rotor, the planned component that will replace Turbine for block propagation. Anza says Rotor has no committed release date. Until it arrives, the key measures are observed finality over time, the gap between Anza’s 150-millisecond target and external readings, and whether the faster settlement holds as usage changes.

Sources and documents

• mainnet announcement — prnewswire.com
• OpenChainBench’s 24-hour measurement through Oct. 23 — openchainbench.com