AdvertiseFee Deal Desk

Company Announcements

Anza Moves Solana (SOL) Alpenglow Upgrade to Public Testnet Aiming for 150ms Finality

Anza moves Solana's Alpenglow upgrade to public testnet, replacing TowerBFT with Votor to cut finality from 12.8s to ~150ms; mainnet set for Sept 28.

Be a creator
September 25, 2026, 04:59 AM UTC6 min readUpdated
AI SummaryAI
  • Anza's Alpenglow upgrade entered public testnet targeting ~150ms finality, down from ~12.8 seconds.
  • Votor replaces TowerBFT, letting validators finalize blocks in one or two voting rounds.
  • Anza's feature gate tracker marked Alpenglow as Pending Testnet Activation as of September 24.
  • Mainnet validator upgrade to Agave 4.3 carries a tentative date of September 28.
gate.com

Alpenglow Reaches Public Testnet

Solana (SOL) has pushed its most significant protocol overhaul of the year a decisive step forward: Alpenglow, the upgrade that dismantles the network’s TowerBFT consensus engine, is now staged on the public testnet. The change targets transaction finality — the point at which a block becomes irreversible — cutting it from roughly 12.8 seconds today to about 150 milliseconds, or less than one-sixth of a second. The Solana ecosystem’s core client developer, Anza, circulated a detailed upgrade guide and migration checklist through the Solana Foundation’s developer channel on September 21, and its feature gate tracker listed Alpenglow as “Pending Testnet Activation” as of September 24, with a tentative mainnet date of September 28 now on the schedule.

Finality is among the most consequential metrics on any chain. Exchanges credit deposits only after it, bridges release funds on it, and merchants treat it as the moment a payment cannot be reversed. A 150-millisecond target would put Solana’s settlement speed on par with — or ahead of — traditional financial rails, a point the Solana (SOL) Foundation stresses when describing point-of-sale use cases. Anza is staging the rollout deliberately: after recommending Agave 4.3, the reference validator software it maintains, to all mainnet validators on September 21, the team plans a phased migration in which operators controlling 10% and then 25% of total SOL staking weight upgrade first, reducing the risk of a synchronized network-wide switch. Alpenglow has already run for more than four months on small dedicated clusters; the public testnet move exposes the code to a far wider, more heterogeneous set of machines and services, using valueless test tokens so developers can debug — and even reset the network — without touching user funds.

Votor, Certificates and Client Risk

The technical heart of Alpenglow is Votor, a new voting protocol that rewrites how consensus is reached. Under TowerBFT today, validators broadcast votes as on-chain transactions, and a block finalizes only after 32 slots of accumulated votes — roughly 12.8 seconds per the Foundation’s own documentation. Votor removes the on-chain vote chain entirely: validators exchange ballots directly with one another, aggregate them into certificates, and finalize a block in one or two voting rounds. The ~150ms figure is a target rather than a fixed constant — it varies with network latency, stake distribution and the number of rounds required.

A second phase will later replace Turbine, the block-propagation system that shreds data into small units for distribution, with Rotor; its timing is not yet scheduled. Crucially, Alpenglow leaves the Solana (SOL) Virtual Machine, transaction format, programs and fees untouched, so wallets and dapps require no user-facing migration. Validator operators, however, must register a BLS public key — a requirement already in force — and upgrade to Agave 4.3; unregistered validators are excluded from consensus participation. Indexers, block explorers, trading infrastructure and block builders that handle block and vote data directly must also adjust their processing. One gap stands out: Jump Crypto’s alternative clients, Firedancer and Frankendancer, do not yet support Alpenglow testing, meaning the first mainnet deployment runs on a single software stack. That echoes April 2026, when Solana shortened its slot time from 400ms to 250ms with Agave leading and Firedancer following; that upgrade — a parameter change, not a consensus overhaul — still produced brief network latency. Alpenglow is deeper surgery, effectively a consensus-layer fork, so the staged 10%/25% rollout is the principal hedge against a network-wide fault. Readers tracking the market in real time can follow live spot and futures prices on Bitget.

What 150ms Finality Unlocks

Update: the rollout has widened beyond the first testnet. Anza announced on September 25 that Alpenglow is now active on Solana's devnet as well, a day after the separate public testnet completed its own transition, and the Solana Foundation's upgrade dashboard lists the feature as live on both networks. Devnet gives application teams a sandbox with valueless tokens to verify their software before mainnet switches. Two caveats from the Foundation's guidance deserve attention: once validator votes leave blocks, network-wide transaction counts will fall even if user activity is unchanged, so data providers have been told to adjust their comparisons, and indexers must keep competing candidate blocks separate to avoid recording a corrupted history. No firm mainnet date has been set — the tentative September 28 window covers feature activations generally, and the 150ms figure remains a simulation-derived target, not yet a demonstrated result under live-market conditions.

Further details on the devnet activation clarify how the two networks divide the rollout's testing burden. Application teams use devnet to verify how their software behaves with Alpenglow's consensus changes before user funds are at stake, while testnet serves validators and infrastructure operators exercising the network software under heavier load. The Foundation's upgrade page confirms both networks run the new system, and notes that Votor can finalize a block in a single voting round when validators representing at least 80% of stake accept it, with a second round as a fallback threshold. Mainnet remains on TowerBFT, the Foundation's status page still lists Alpenglow as inactive there, and the schedule does not identify September 28 as an Alpenglow-specific date.

(as of 07:13 UTC) Our desk’s reading: Alpenglow is less a speed stunt than a repositioning — Solana is trading raw-throughput headlines for deterministic, sub-second settlement that payment rails and institutional desks can actually underwrite, at a moment when Solana DEX activity already processed 208 million trades in a single week by our prior count. The feature gate schedule leaves node operators with concrete deadlines: BLS key registration enforced, Agave 4.3 mandatory for the testnet migration, and a tentative September 28 mainnet date. Institutional staking exposure — including ARK Invest’s recent purchase of 3iQ’s Solana staking ETF — raises the cost of any consensus failure, which is precisely why the staged rollout, not the 150ms headline, is the number to watch. Traders mapping upside scenarios such as Peter Brandt’s 106% Solana target should weigh client-diversity risk just as closely.

Primary sources

COINOTAG's editorial and research desk.

AI-Assisted

AI-generated, AI-reviewed, under COINOTAG editorial oversight.