Solana's clock finally halved: 200ms slots are live, and an epoch now lasts about a day

On 9 October 2026, Solana mainnet entered epoch 1053 and its clock ticked over to a 200-millisecond slot. The chain's own timestamp puts the first slot of that epoch, slot 454,896,000, at about 14:41 UTC. Anza confirmed the switch minutes later. A slot is the short turn one validator, the leader, gets to produce a block. That turn is now half the original 400ms, so the network schedules twice as many slots every second as it did at launch.

An epoch is the other clock, and it is the one you feel on a calendar. An epoch is always 432,000 slots. Count them at 400ms and you get 48 hours. Count them at 200ms and you get 24 hours, about a day. Staking rewards pay out once per epoch, so they now arrive about daily. At the 250ms step, from 18 September until 9 October, an epoch was about 30 hours.

This is the last of four cuts under SIMD-0525: 400, then 350, then 300, then the 250ms step in September, and now 200. The slot-time roadmap that Agave 4.2 shipped behind feature gates is complete.

Why a shorter slot matters

Picture the network as a relay. Each leader holds the baton for four slots in a row. Each of those slots is a turn to produce a block. At the original 400ms, a turn lasted 1.6 seconds. At 250ms it was one second. At 200ms it lasts 0.8 seconds. Shorter turns mean the chain you see is fresher, and one leader has less time to delay, reorder, or quietly leave out transactions before the baton moves on.

The four cuts, each one taking effect at the start of an epoch:

• 350ms – 21 August 2026, epoch 1020
• 300ms – 28 August 2026, epoch 1024
• 250ms – 18 September 2026, epoch 1037
• 200ms – 9 October 2026, epoch 1053

Every step followed the same two-beat rhythm written into SIMD-0525: the switch flips on at one epoch boundary, and the shorter slot starts governing blocks at the next. For this final step the switch turned on at epoch 1052 and the 200ms slots began at 1053. That one-epoch pause gives every validator notice before the network starts enforcing the smaller per-slot limits. The Solana Foundation tracker records the same dates. "Final" means final for this proposal. Other upgrades keep their own calendars.

Twice the beats, the same amount of work

Here is the part worth getting right. Five slots a second sounds like more throughput. It is the same throughput, cut into smaller pieces. SIMD-0525 shrinks the compute and data allowed in each block by the same proportion as the slot. Mainnet's block limit at the 400ms baseline has been 100 million compute units since July; scaled to 200ms, one block holds 50 million. Half the slot, half the block, twice as many blocks: the work the network can do each second comes out about where it was. The same applies to data: a block carries at most 16,384 data shreds now, against 32,768 at 400ms.

That wait is latency, and it is what you gain. Wallets, exchanges, and trading bots see the latest state sooner. A confirmation still counts the same number of slots, and each slot is shorter, so it lands sooner on the clock. Alpenglow is a separate upgrade, and it is still off. Its target is finality of about 150ms: the moment a block can no longer be undone. The 200ms figure is the slot, the turn in which a leader produces a block.

One more clock gets shorter. A transaction carries a recent blockhash, a stamp that proves it was built against the live chain. That stamp stays valid for 150 blocks. With a block every 200ms, that is about 30 seconds, against 37.5 seconds at 250ms and 60 seconds at 400ms. Hardware wallets, multisig flows, and anything that waits for a person to press approve now have less slack before the stamp goes stale.

Epoch length comes from the same arithmetic. SIMD-0525's table gives 24 hours at 200ms, 30 hours at 250ms, and 48 hours at 400ms, and it says plainly that real epochs can run a little off those figures. Anza's post puts it as an epoch "about a day instead of two". The first full epoch on this clock will show the exact length. Over a year, the amount of new SOL stays the same: SIMD-0525 raises slots_per_year so annual inflation holds steady.

What it means for…

Validators

A leader's turn is 0.8 seconds. Replay, voting, and block packing have less real time than they did at 250ms, and far less than at 400ms. Epochs come around about once a day, so reward cycles and epoch-boundary chores do too. The switch itself was a feature gate already in the client; operators on an older client show up in skip statistics. Block skip rate was the check before every step of this rollout. Keep an eye on it at solana.com/200ms.

Delegators

Your stake stays where it is. Rewards follow the same annual rate and now land about once a day instead of every two. Favour validators who publish their client version and keep skip rates low on the tighter clock. The apps you use feel snappier. The network's total capacity per second stays about where it was.

Builders

Read getBlockTime or the current slot duration whenever you turn slots into clock time. Blockhash expiry is about 30 seconds, so retry logic and any signing flow that waits on a person should plan for less time. The limit on a single transaction stays as it is; the block it fits into holds 50 million CUs. Indexers keep the same block format and take in about twice as many blocks per day as on the original 400ms clock. Design screens for updates that arrive a bit sooner.

Everyone else

If you hold or send SOL, this is something you feel rather than do. Transfers and swaps should confirm a little sooner than they did at 250ms, and apps should look less "one beat behind". Two clocks are worth keeping straight: a slot of 200ms, and an epoch of about a day.

Sources: Anza – 200ms slots live, 9 October 2026 · SIMD-0525 · Solana Foundation – Reduced slot times · Live 200ms board