For about 400 milliseconds this afternoon, Solana will stop landing user transactions. That isn’t a failure.
It’s the scheduled handoff between two consensus systems, and it’s probably the only “400 ms” figure that matters today.
Many people heading into the Alpenglow launch still treat 400 ms block times and 12.8-second finality as the bar to beat. Those are the old chain’s numbers. They describe TowerBFT, the system Alpenglow is replacing, so Alpenglow shouldn’t be checked against them. If you go through Solana’s own documentation, you’ll find the right benchmarks before the first post-launch numbers arrive.
Alpenglow’s Switch Is Scheduled for Slot 459,653,000
According to Solana’s official Alpenglow page, the upgrade’s feature gate activates at the start of epoch 1064 on Wednesday, October 21, at about 13:20 UTC. The Solana Foundation says to treat that time as an estimate and to watch the epoch instead.
The new consensus doesn’t take over the moment the epoch begins. TowerBFT keeps running for the first 5,000 slots, and Alpenglow takes control at slot 459,653,000. Anza, the team behind the main Agave client, put that switch at around 14:00 UTC.
That time isn’t fixed. Solana Compass calculated that with about 375,000 slots left before the switch, each millisecond of drift in average slot time moves the boundary by roughly six minutes. Using epoch 1062’s pace, it estimated the switch at about 14:02 UTC.
The transition itself has a cost. Solana Compass reports a gap of roughly two slots, about 400 ms, during which blocks carry only votes and user transactions don’t land. So the 400 ms everyone is watching for will turn up, just as a pause rather than a block time.
The 400 ms and 12.8 Second Figures Describe the Old Solana
This is where much of today’s coverage gets it wrong.
Block time first. Solana’s page says slots moved from 400 ms to 200 ms in a separate “reduced slot times” upgrade, which finished this month. That change isn’t part of Alpenglow. It was already running before today. Solana Compass measured epoch 1062 averaging 217.7 ms per slot over 26.1 hours. Checking Alpenglow against 400 ms would mean grading it against a number the network had already beaten before the switch.
Then finality. The 12.8 seconds is TowerBFT’s figure. Under that system, a block became final only after enough votes built up over 32 slots. Alpenglow’s stated target is about 150 ms, which Solana describes as 99% faster than TowerBFT.
Solana is careful about that 150 ms figure, and I think readers should be too. The Foundation calls it an expectation, not a constant, and says one-round finality comes in lower. Galaxy’s engineers, quoted by Solana Compass, call it a design target rather than a measured result. As I noted when looking at how CertiK verifies a TPS claim, a target is a claim until someone measures it on the live network.
So the right checks for today are:
- Slot time: about 200 ms (already the case before today).
- Finality: about 150 ms, down from 12.8 seconds.
What Alpenglow Actually Changes Under the Hood
Alpenglow is Solana’s first replacement of its core consensus since TowerBFT, and today’s launch only covers part of it.
The first phase, Votor, ships in Agave 4.3. Validators now send votes directly to each other and combine them into certificates, instead of submitting votes as on-chain transactions. Per Solana’s page, votes are no longer transactions and no longer appear in blocks. A block can finalize in one round if at least 80% of stake notarizes it. Otherwise, a second round uses 60% thresholds.
Proof of History no longer paces block production. Execution stays the same: the SVM, programs, transactions and fees don’t change.
The second phase, Rotor, will replace Turbine, Solana’s block propagation layer. Solana lists it as “not yet scheduled.” If someone credits today’s numbers to “Alpenglow” as a whole, half of the design hasn’t shipped yet.
The security trade-off is also worth knowing. TowerBFT needed more than two-thirds of stake for confirmation. Alpenglow tolerates 20% malicious stake plus 20% offline stake. Solana Compass notes that slashing for double voting is defined but not enabled in this first version.
Validators also pay differently. TowerBFT voting cost about 1 SOL per epoch in vote transaction fees. Alpenglow replaces that with a flat 0.8 SOL per epoch Validator Admission Ticket (SIMD-0357), which is burned. A validator that can’t pay doesn’t vote or produce blocks in the next epoch.
The 80% Problem in the Readiness Numbers
In my view, this is the detail most likely to affect today’s real numbers.
Alpenglow’s fastest finality depends on the one-round fast path, which needs at least 80% of stake to notarize a block. On the morning of October 20, a tracker run by Pumpkin’s Pool, cited by SolanaFloor and reported by Solana Compass, flagged 151 validators holding 20.3% of stake as unprepared. The day before, that figure was 45%.
The math is uncomfortable. If 20.3% of stake really couldn’t vote, only 79.7% would be left, which is just short of the fast-path threshold. In that case, blocks would need the slower second round, and finality would land above the headline figure.
I expect that number to drop before the switch, since operators have been upgrading quickly. But it’s the main reason I wouldn’t trust any finality figure from the first hour without also checking how much stake is actually voting. Validators without a registered BLS key, or that miss the required client version, stop voting entirely. Their delegators also stop earning inflation rewards.
What the Test Cluster Shows, and What It Can’t
There’s already one public reference point. Solana’s page links to a community cluster explorer run by Valid Blocks, where Alpenglow has been running on Agave 4.3.2.
When I checked this morning, it showed an average finalization time of 131 ms over the past hour, ranging from 111 ms to 156 ms. Average slot time was 215 ms. That’s inside Solana’s target range.
That cluster had 62 active validators and about 10.2 million SOL staked. Mainnet is far larger, more spread out geographically, and runs under real load. Test-cluster results are a strong sign, but they aren’t proof of how mainnet will perform.
How to Check the Real Numbers Yourself
You don’t need to wait for a press release to know whether the switch happened. Solana’s getAgGenesisCert RPC method returns null while the cluster is still on TowerBFT and returns a genesis certificate once Alpenglow is live. Nodes older than Agave 4.3 return a “Method not found” error, which shouldn’t be confused with null.
Once the certificate shows up, here’s what I’ll be watching over the first few epochs:
- Average slot time, compared with the roughly 217 ms baseline from epoch 1062, not with 400 ms.
- Finality, compared with the roughly 150 ms target, with separate figures for one-round and two-round blocks.
- Voting stake, to see whether the 80% fast path is regularly reachable.
One caveat for developers: according to Anza’s operator notes, the Clock sysvar’s unix_timestamp now estimates when the parent block ended, not when the current block began, and should be treated as approximate. Any timing tool that reads block timestamps may need adjusting.
If Alpenglow delivers, Solana goes from finality you wait for to finality you’d barely notice. I’ll believe the 150 ms figure once mainnet consistently shows it.
Disclosure: This is not trading or investment advice. Always do your research before buying any cryptocurrency or investing in any services. Follow us on X @nulltxnews

