Solana Alpenglow Is Live on Mainnet as Anza Retires TowerBFT at Slot 459,653,000

Solana Alpenglow is live on mainnet-beta. Anza announced at 13:54 UTC on 21 October 2026 that the network's new consensus protocol had taken over, retiring TowerBFT, the voting system Solana has run since its 2020 launch. Anza's post put the change at "finality reduced from 12.8 seconds to 150 milliseconds". The handoff came at slot 459,653,000 in epoch 1064, the boundary we described in our pending-activation report on 20 October.

Finality is the point at which a transaction can no longer be reversed. Alpenglow's first component, Votor, changes how validators agree on blocks, and it removes the vote transactions that made up most of Solana's on-chain traffic. The 150ms figure is Anza's own. No independent post-switch measurement had been published when this article went out.

Solana Alpenglow went live at slot 459,653,000, about 13:54 UTC

The switch landed a few minutes before 14:00 UTC on Wednesday 21 October. Anza's announcement is timestamped 13:54:56 UTC, and Cointelegraph reported that Alpenglow "went live on mainnet around 1:54 pm UTC". Solana Compass's epoch data fits that timing. It puts the start of epoch 1064 at about 13:36 UTC, and the migration design waits 5,000 slots after the epoch boundary before the handoff, roughly 17 to 18 minutes at Solana's current slot pace.

The official Solana account followed at 13:59 UTC, calling it "the biggest consensus upgrade in Solana's history, built by @anza_xyz." Jacob Creech, whom Decrypt identifies as the Solana Foundation's technology vice president, summed up the result at 13:56 UTC: "Vote transactions -> 0, Finality -> 150ms, TowerBFT -> Gone".

The migration could not complete without broad participation. Bitcoin.com reported that the process "required participation from validators representing at least 82% of staked SOL" before the transition could finish. The mechanics, from vote-only blocks to the genesis certificate that starts Alpenglow, are laid out in our report on the mainnet date. Mainnet followed the same path as the testnet handoff on 24 September and the devnet handoff a day later.

How Votor replaces TowerBFT: validator votes move off-chain

Under TowerBFT, every validator voted on blocks by sending a transaction on-chain, and a block took many rounds of those votes to reach finality. Votor has validators exchange votes directly and bundle them into certificates, so a block can be finalized in one or two rounds of voting.

Galaxy, which ran validators on the Alpenglow test cluster, describes the two paths: if 80% of stake votes for a block, "it finalizes in a single round"; with lower participation, a second round finalizes it once separate 60% certificates exist. Roger Wattenhofer, Anza's head of research, put the fast path more plainly to Decrypt: "If you have 80% of the votes say yes, then it's immediately done. You immediately know that the block is final."

Removing votes from blocks changes what the chain carries. In the week before the switch, 14 to 20 October, vote transactions made up about 62% of everything Solana processed, roughly 263 million a day against 145 million to 175 million non-vote transactions, according to Solana Compass network data. Those vote transactions no longer exist under Alpenglow. Network-wide transaction totals from 21 October onward count user activity only and are not comparable with the totals before it.

The 150ms finality figure is Anza's, and independent measurement is pending

Anza and the Solana account both give the same headline: about 12.8 seconds before, about 150 milliseconds after. Anza's press release says Alpenglow "is designed to deliver" finality "in roughly 150 milliseconds", and Bitcoin.com reports the Solana Foundation's own caveat that "actual performance will depend on network conditions". Max Resnick, Anza's lead economist, told Decrypt that users should first notice "a reduction in confirmation latency to approximately 150-200ms."

The 12.8-second baseline also deserves a footnote. Galaxy puts TowerBFT finality at "about 6 seconds at today's 200ms slots", because Solana's slot time reached 200ms on 9 October after stepping down from 400ms. The 12.8-second comparison reflects the longer slots Solana ran for most of its history. Against either baseline the claimed improvement is large. What the network has not yet shown is a measured finality figure from mainnet, which will come from dashboards and RPC providers over the coming days.

In the release, Anza chief executive Brennan Watt called Alpenglow "the largest, most ambitious blockchain upgrade since 'The Merge'" and said Anza took it from a white paper to mainnet in under 18 months.

Watt's comparison is to Ethereum's 2022 switch from proof-of-work to proof-of-stake. Solana validators approved the Alpenglow proposal in a September 2025 governance vote that Decrypt reports passed with 98% in favour.

Solana non-vote transactions ran at normal levels through the consensus switch

The handoff did not interrupt user activity. Solana Compass's hourly data shows about 2,150 non-vote transactions per second in the 13:00 UTC hour, the hour in which the switch landed, and about 2,500 in the 14:00 UTC hour, the first full hour on Alpenglow. The same hours on 20 October ran at about 1,990 and 2,640. Non-vote transactions never included validator votes, which makes them the like-for-like series across the boundary.

Solana validators now pay a 0.8 SOL admission ticket instead of vote fees

For the operators who run the network, the biggest economic change is the end of vote fees. Under TowerBFT, each validator paid transaction fees on every vote it cast. The Solana Foundation's upgrade guide says Alpenglow instead charges "each voting validator a flat 0.8 SOL per epoch" through a Validator Admission Ticket, deducted from the vote account. The first ticket was charged at the start of epoch 1064 and pays for epoch 1065.

The Foundation guide required every operator in Solana's active validator set to run version 4.3.2 of an Agave-based client (Agave, Jito JTO$0.484+2.1%-Solana or Harmonic Salsa) for the migration, the release Anza named as recommended on 16 October. Firedancer and Frankendancer were not supported during the switch. Galaxy wrote that Firedancer's 26.10 release supports Alpenglow, so operators can return to it after the transition, while "Frankendancer will not be supported once Alpenglow is active."

Readiness was the open question going into the week. On 19 October, Pumpkin's Pool's Alpenglow tracker flagged 45% of stake as not ready, much of it on earlier 4.3 builds. No count of ready stake at the moment of the switch has been published; the completed handoff confirms only that validators holding at least 82% of stake took part.

What changes for Solana apps, RPC providers and indexers: bank_id and block footers

For most developers nothing changes. The Foundation guide says "the SVM, transactions, programs, and fees stay the same", so apps that send and sign transactions need no update.

Services that read blocks in real time do have work to do. Under Alpenglow "a slot can carry more than one candidate block", so block-stream consumers have to track a new bank_id field and keep only the bank that reaches confirmation, as covered in Triton One's integrator guide earlier this month. Triton One said at 14:15 UTC that it supports the new block footer across its gRPC, Fumarole and JSON-RPC services.

Next on the roadmap: Rotor, geographic leader rotation and Quantumglow

Votor is phase one. Rotor, the second phase, is designed to replace the way blocks propagate across the network, and the Foundation guide says it "is planned for a later release once Votor is adopted, and is not yet scheduled." Anza's release lists Rotor alongside geographically aware leader rotation, which Anza researchers have proposed in SIMD-0674 and SIMD-0675, and Quantumglow, a post-quantum version of Alpenglow, as items with "no committed date yet."

Decrypt reports that developers on Jump Crypto's Firedancer team have proposed scrapping the per-block compute cap once Alpenglow is live, and Cointelegraph reports that Agave 4.5, planned for December 2026, would raise the number of accounts a versioned transaction can reference from 64 to 96.

For now, the verified result is the handoff itself. Solana replaced the protocol its validators use to agree on every block, and user traffic ran through the boundary at normal levels. Anza's 150ms figure is the next thing mainnet has to show in measured numbers.