Solana's mainnet is now producing blocks on a 350-millisecond target, down from 400ms, after the network activated the change at epoch 1020 on Aug. 21. The shift, which required a one-epoch delay under the SIMD-0525 proposal, marks the first step in a roadmap that aims for 200ms eventually.
Why the one-epoch wait
The feature gate went live at slot 440,208,000, which is epoch 1019. But the actual change to block time didn't take effect until the next epoch, 1020, because SIMD-0525 requires a full epoch to let the network settle before a new timing target kicks in. That's the same mechanism Solana will use for the next stages.
Brennan Watt, CEO of Anza, said the 300ms stage is intended to become effective at epoch 1024, which should arrive around Aug. 28. That hasn't happened yet, as of Aug. 26. The network can pause between stages if skipped-block rates climb, meaning the 350ms run is also a live test for how much tighter the chain can go without breaking.
What the numbers show
Telemetry from Trillium measured a slot-weighted mean of 365.4ms in the post-change epoch 1021. For comparison, the pre-change epoch 1015 ran at 420.7ms. So the actual average is already below the old target, though still above the new one.
Skipped slots fell sharply. In epoch 1015, there were 1,890 skipped slots, or 0.438% of the total. In epoch 1021, that dropped to 331, which is 0.077%. Early indicators point to stability, not degradation, after the change.
Faster slots, shorter leader window
Block time directly affects how long a leader holds the slot window. With four slots per leader, the window shrinks from 1.6 seconds to 1.4 seconds. That's a 12.5% reduction in how long each leader controls the network's output.
The proposal keeps four slots per leader and 432,000 slots per epoch. But the per-slot compute, account-write, vote, data, and shred limits all scale down proportionally to the shorter target. So the network doesn't just run faster — it also does less per slot to keep the overall throughput steady.
The roadmap calls for 300ms, then 250ms, and finally 200ms. The 300ms stage is still pending, and it's the one that Watt says should arrive at epoch 1024. But the network won't move forward automatically — if skipped-block rates start climbing, the pause kicks in and the next stage waits.
The real test is whether the 350ms stage holds up over several more epochs. If it does, the jump to 300ms will come fast. If not, the pause will give validators time to catch up. Either way, the next decision point is at epoch 1024, when the 300ms gate is scheduled to be active.
