Solana Alpenglow Reaches Devnet: The 150ms Finality Era Moves One Step Closer

Solana’s Alpenglow Reaches Devnet: The 150ms Finality Era Moves One Step Closer

Solana’s most consequential consensus overhaul in years has cleared its latest milestone. The Alpenglow upgrade is now live on the network’s public developer network, giving application teams their first hands-on look at a system designed to cut transaction finality from roughly 12.8 seconds to about 150 milliseconds.

Core developer firm Anza announced the devnet switch on Sept. 25, one day after Solana’s testnet completed its own transition. According to the Solana Foundation’s upgrade page, Alpenglow is now active on both devnet and testnet — but not on mainnet, which continues to run the network’s long-standing TowerBFT consensus. That means the headline 150-millisecond figure remains a target validated in testing and simulations rather than a finality time available to users transacting today.

Why finality, not just speed, matters

The two staging networks serve distinct roles in the rollout. Devnet lets application teams test their software against the new consensus using tokens that carry no real value, while testnet gives validators and infrastructure operators a place to run the network software under more demanding conditions.

Finality is the point at which a blockchain has agreed on a transaction strongly enough that it can no longer be reversed under the network’s consensus rules. For everyday users the difference between 12.8 seconds and 150 milliseconds may sound academic, but it changes decisions at institutional scale. A United States exchange deciding when to credit a Solana deposit, or a payment provider deciding when to treat a merchant’s sale as complete, currently has to build waiting time into its flows. Faster finality compresses those windows — though each service may still apply its own checks before releasing funds.

How Votor replaces TowerBFT voting

Under TowerBFT, validators submit votes as transactions that appear inside blocks. Enough votes must accumulate across 32 slots before a block becomes final, a process that takes about 12.8 seconds, according to the Foundation.

Alpenglow’s first phase, called Votor, takes a different approach: validators send votes directly to one another instead of embedding them in blocks. A block can reach finality after a single voting round if validators representing at least 80 percent of stake vote to accept it, with a second round providing another path when the first does not meet the threshold.

The Foundation is careful to separate finality from block production. In September, Solana reduced its target slot time from 300 milliseconds to 250 milliseconds, with a further reduction to 200 milliseconds planned under a separate upgrade. Shorter slots change how often the network can produce blocks; Alpenglow changes how validators agree that a block is final. The two improvements are complementary pieces of the same speed push.

What developers and data providers must change

For an application that only sends transactions and reads account balances, the Foundation says Alpenglow requires no migration at all. Transaction execution, fees and the formats used to submit transactions remain unchanged under the consensus upgrade.

Services that build transaction histories face more work. Alpenglow can expose competing candidate blocks for the same slot before the network selects one, and the Foundation has warned data providers to keep those candidates separate and retain only the block that reaches confirmation. Merging transactions from different candidates could leave an explorer or analytics service with an incorrect record.

Validator votes will also disappear from blocks entirely, since they will no longer be submitted as transactions. That has a subtle consequence: any chart counting both user transactions and validator votes will show a lower total after activation even if users make exactly the same number of payments and trades. The Foundation has advised data providers to reset comparisons and alerts built on the old figures. Services that read validator participation from vote transactions will find that information moved to certificates attached to block data, and operators relying on Solana’s Geyser or gRPC data streams must account for new identifiers that distinguish candidate blocks within a slot.

Mainnet still waits for a date

Solana’s earlier Alpenglow roadmap tied the proposed mainnet rollout to the Agave 4.3 software release and an October target. Neither the testnet transition nor the devnet activation sets a confirmed date for the live-network switch. Anza’s software schedule tentatively allows mainnet feature activations to resume on Sept. 28, but the schedule does not identify that day as Alpenglow’s activation date, and the Foundation’s status page still lists the upgrade as inactive on mainnet.

The Foundation describes Votor as only the first phase of Alpenglow. A later phase, Rotor, is planned to replace the system used to spread blocks across the network. For now, the devnet milestone gives Solana’s sprawling application ecosystem — from decentralized exchanges to payment rails — the runway to prepare for a consensus change that could redefine what the network is capable of.

Market snapshot at press time (Sept. 27, 2026, 13:50 UTC): SOL at 122.89 USD, up 1.85 percent in 24 hours; BTC at 84,859 USD, up 1.09 percent; ETH at 2,705.03 USD, up 0.75 percent.