Alpenglow Goes Live on Solana's Devnet With a 150ms Finality Target

Solana's consensus rewrite took over the network's developer cluster on Friday evening. Anza, the firm that maintains Solana's core validator software, switched devnet to Alpenglow at slot 504,148,999 at 20:01 UTC, retiring the old TowerBFT voting system on that cluster. Devnet is the second public test network to make the change. The target is finality of about 150 milliseconds, against roughly 12.8 seconds now, finality being the point at which a transaction can no longer be undone. Nothing published this week sets a mainnet date.

What Votor changes

Alpenglow retires TowerBFT for Votor, in which validators send their votes straight to one another and aggregate them into certificates, instead of writing each vote into a block as a transaction. CoinDesk, filing at 01:00 New York time on Saturday, put the mechanic plainly: validators "will exchange votes directly instead of recording those votes as transactions inside blocks." Phemex says a block can settle after one or two rounds. Coinpaprika describes two routes through the scheme, a fast one needing 80% of stake and a fallback at 60%, and no other account in this sweep carries those thresholds.

Execution and fees are untouched. Fault tolerance goes up, to 40% of validators failing against about 33% today, in cases where the failures are not adversarial; coinpaprika and an August schedule note from Solana Compass both give that pair. Ordinary wallets and applications need no migration. The work sits alongside the slot-time cuts this desk has covered, including the move to 250ms slots at epoch 1037, but it is a separate track.

The transaction count is about to fall

One consequence lands on anyone who publishes Solana numbers. Votes stop being transactions, so the headline count drops while users do exactly as much as before. The Solana Foundation has told data providers to adjust their metrics for that, on Phemex's account of the guidance, and told infrastructure services to hold competing candidate blocks apart until a block is final. Solana Compass, writing on Friday night, puts vote traffic at 54% to 59% of daily activity on mainnet-beta before the switch. That range appears in no other account here, so the direction is confirmed and the figure is one reading.

Two accounts of when testnet switched

The devnet timing is exact. The feature gate flipped at epoch 1167, about 19:40 UTC, and a 5,000-slot countdown ran roughly 17 minutes before Alpenglow took control. Testnet is where the record splits. Cryptonomist and coinpaprika both date that activation to 23 September; Solana Compass says devnet followed testnet by one day, which puts it on the 24th. Neither side acknowledges the other, and we could not establish which date supersedes which.

Why 28 September is not the date

A date has been circulating and it is the wrong one. 28 September is when Anza resumes working through queued feature gates on mainnet, in the language of its own Agave 4.3 release schedule. It is not Alpenglow's activation. CoinDesk calls the date tentative, Phemex says outright that it is not guaranteed as the deployment date, and the schedule itself carries the correction. Validators have been on that release since general adoption was recommended on 21 September, the branch this desk covered when Firedancer v1.1.0 reached testnet.

Developers should "test your programs and integrations now during our observation period leading up to mainnet-beta migration," Anza said.

Whatever mainnet date arrives, some of the work before it is administrative. One account of the release schedule says validators have to register a BLS public key on chain before the feature gate opens, and that those who skip it are left out of the consensus set. That requirement sits in a single source here. CoinDesk makes the other caveat that matters: 150 milliseconds is a simulation figure, not a number measured under live market conditions, and wallet processing and exchange deposit checks sit on top of whatever the network itself does. Devnet is where that gap starts to show.