SIMD-0525 shortens each epoch to 42 hours and rotates validators every 1.4 seconds. The change cuts confirmation latency without adding throughput; three more 50ms cuts are queued.
Solana's mainnet cut its target slot time from 400 milliseconds to 350 at the start of epoch 1020 on Friday, the first such reduction since the network launched in 2020. Every epoch on Solana now runs closer to 42 hours instead of 48, and each leader window is 1.4 seconds long instead of 1.6.
The change is the first stage of SIMD-0525, a Solana Improvement Document merged in May that lays out four sequential 50-millisecond cuts to bring slot time down to 200ms. Each stage requires two-thirds of staked validators to endorse it, and the rollout can be paused if too many validators fail to produce blocks in their assigned slots.
The upgrade shipped inside Agave 4.2, the validator client release Anza, the Solana-focused firm that inherited most of the original Solana Labs engineering team, began activating the week of August 17. The feature gate went live at slot 440,208,000, the opening slot of epoch 1019, with a mandatory one-epoch delay before the new timing took effect. The 350-millisecond target went live when epoch 1020 opened on Friday.
None of this changes how much work Solana can do per second. Validators handle slots more often but each slot carries less work; the network's overall compute capacity is not affected by the timing change. That capacity was separately raised to 100 million compute units per block in July 2025. What the slot cut buys is latency: shorter confirmation windows, faster leader rotation, and less time for any single validator to sit on transactions before its neighbour takes over.
The Solana Foundation's upgrade documentation frames the change through the leader-window mechanic. A shorter leader window makes ordering games harder to run. A validator that wants to delay a swap, front-run a liquidation, or rearrange a block of transactions has less time to do it before the next validator is holding the pen. That matters more on Solana than on most chains, because Solana's leader schedule is deterministic and publicly known — a validator that abuses its slot knows exactly when its window closes.
The move also compresses Solana's epoch structure. Each epoch still contains 432,000 slots. At 350 milliseconds a slot, that comes to just over 42 hours, down from about 48. Stake rewards, vote credits, and every other epoch-scheduled process now cycle six hours faster. For anyone building governance systems, dApps that track epochs, or MEV infrastructure that assumes 48-hour cycles, the change requires updated timing assumptions.
The Block, running a point-in-time measurement for its coverage, timed a 1,000-slot window before the change at 415 seconds and a comparable window inside epoch 1020 at 368 seconds. The overhead, around 18 milliseconds per slot beyond the 350-millisecond target, reflects the network's real-world block time, which has always sat slightly above the target because leaders occasionally miss their slots.
Firedancer, Jump Crypto's C-based validator client, has been on Solana mainnet since December 2025, giving the network its second production client alongside Agave. Client diversity matters here because a coordinated timing change of this scale historically stresses less-mature clients hardest. So far there is no reporting that Firedancer nodes have struggled with the shorter target.
The next planned cut is to 300 milliseconds, but Solana developers have not set an activation date. The Solana Foundation's guidance is that they will monitor the network's behaviour at 350 milliseconds before proceeding. If leader skip rates or fork rates rise, the rollout pauses.
Slot time is not the same as finality. Full finality, the point at which a Solana block is treated as economically irreversible, still takes roughly 12.8 seconds today, unchanged by this upgrade. Anza's Alpenglow overhaul, still in development, aims to bring finality down to about 150 milliseconds; that is a separate consensus rewrite, not a slot-time change. A user experience of near-instant transactions on Solana, as a networked product rather than an aspirational target, depends on Alpenglow shipping.
For Solana, which spent much of 2025 rebuilding after the memecoin-driven fee collapse, the slot cut is a concrete engineering win in a year where most of the marketing has been about the spot ETFs that launched earlier and about restaking primitives like the Jito vaults. Faster confirmations do not fix any of the network's economic problems, but they do give protocols building on Solana a technical claim they cannot make against Ethereum's own base layer.
The final 200-millisecond target sits two more stages out, and the schedule for reaching it depends entirely on how the current 350ms behaves in production. If it holds, the next Solana users notice will move a little faster than the ones this week did.
