Twelve point eight seconds. That’s roughly how long you wait today before a Solana transaction is final and can’t be reversed. Alpenglow, the upgrade meant to cut that to about 150 milliseconds, is now running on Solana’s developer network.
That puts it on its second public test network. It still isn’t on the blockchain that moves real money, and nobody has said when it will be.
Two networks, one day apart
Anza, the company that builds Solana’s core software, announced the devnet switch on Sept. 25. The separate testnet had finished its own transition a day earlier. The Solana Foundation’s upgrade page now lists Alpenglow as active on both.
The two networks do different jobs. Devnet is where application teams check their code using tokens with no real value. Testnet is used mainly to stress-test the network software and how validators operate.
For developers, the devnet milestone is the important one. It’s the first time teams can run their own apps against the new behavior before it goes live.
Votes move out of the blocks
Alpenglow changes how validators, the computers that check transactions, agree on what the network has recorded. Under the current design, their votes are written as transactions inside blocks, the batches of transactions added to the chain. With Alpenglow, validators exchange votes directly with each other instead.
Agreement can then take one or two voting rounds. That’s where the 150-millisecond target comes from.
The practical payoff is simple. An exchange could release a deposit sooner. A payment app could tell a merchant a sale is complete without leaving a customer standing at the counter.
Solana’s charts are about to shrink
There’s a side effect that could confuse anyone who watches on-chain numbers. Once votes are no longer transactions, any transaction total that counts validator votes will drop, even if people send just as many payments and trades.
The Foundation has told data providers to adjust their comparisons. If you see a headline about Solana activity collapsing after launch, check which metric it uses before you believe it.
Services that build transaction histories have more work to do. They need to keep competing candidate blocks separate until the network picks one. If they mix the contents, the record they produce could be wrong.
Most apps won’t need to change anything
Apps that only send transactions and read account balances need no migration, according to the Foundation’s guide. That covers a lot of basic wallets and payment tools.
Indexers, explorers and analytics dashboards are a different story. They’re the ones who should be testing on devnet right now.
150 milliseconds is still a simulation
Here’s my main caveat. The 150-millisecond figure is a target based on simulations. It hasn’t been shown under live-market conditions, where real traffic and real validators behave less neatly than a model.
Even if the network hits that number, what you experience won’t be 150 milliseconds. Wallets take time to process transactions, and exchanges run their own deposit checks. Either can add more waiting on top of consensus.
There’s also no firm launch date for the live network. Anza’s software schedule tentatively allows feature activations to resume Sept. 28, but it doesn’t identify that date as Alpenglow’s launch. If you run an indexer or a block explorer, don’t wait for a date to be announced. Point your pipeline at devnet this week and see what breaks while the tokens are still worthless.
