Around 12.8 seconds is what it takes right now for a Solana transaction to become final and irreversible. Alpenglow, the upgrade designed to shrink that wait to roughly 150 milliseconds, has now gone live on Solana’s developer network.
That makes devnet the second public test network to run it. The upgrade has still not reached the blockchain that handles real money, and no one has announced a date for that.
A day apart on two networks
On Sept. 25, Anza, which develops Solana's core software, said the devnet had switched over. The standalone testnet had completed its own changeover the previous day. According to the Solana Foundation’s upgrade page, Alpenglow is now active on both.
Each network serves a different purpose. Application teams use devnet to test their code with tokens that carry no real value. Testnet mostly exists to put the network software and validator operations under stress.
The devnet step is the one developers should care about most. For the first time, teams can try their own apps against the new behavior ahead of the live launch.
Validator votes leave the blocks
Alpenglow reworks how validators, the machines that verify transactions, reach agreement on what the network has recorded. Today, those votes are recorded as transactions within blocks, the bundles of transactions appended to the chain. Under Alpenglow, validators instead send their votes straight to one another.
Reaching agreement then takes either one or two rounds of voting, which is the basis for the 150-millisecond goal.
The real-world benefit is easy to see. Exchanges could free up deposits faster. A payment app could confirm to a merchant that a purchase has gone through without keeping the customer waiting at the till.
Expect Solana’s activity charts to drop
One consequence may trip up people who follow on-chain data. When votes stop being transactions, any transaction count that includes validator votes will fall, even if users send exactly as many payments and trades as before.
The Foundation has asked data providers to adjust their comparisons accordingly. So if a headline after launch claims Solana activity has collapsed, look at which metric it relies on before taking it at face value.
Services that assemble transaction histories face extra work. They must keep rival candidate blocks apart until the network settles on one. Blending their contents could produce an inaccurate record.
Most apps can stay as they are
The Foundation’s guide says apps that simply send transactions and check account balances don’t need any migration. That includes many basic wallets and payment tools.
Indexers, explorers and analytics dashboards are another matter. Those are the teams that ought to be testing on devnet now.
The 150-millisecond number is still simulated
My biggest reservation is this: the 150-millisecond figure is a target drawn from simulations. It hasn’t been demonstrated under live-market conditions, where actual traffic and actual validators are messier than any model.
And even if the network reaches that speed, users won’t feel 150 milliseconds. Wallets need time to handle transactions, and exchanges carry out their own deposit checks. Both can stack extra delay on top of consensus.
There’s no confirmed launch date for the live network either. Anza’s software schedule provisionally lets feature activations restart on Sept. 28, but it does not name that as Alpenglow’s launch date. If you operate an indexer or block explorer, don’t hold out for an announcement. Aim your pipeline at devnet this week and find out what breaks while the tokens are still worthless.