Solana's real speed: block time and transactions per second
Speed claims about Solana range over two orders of magnitude, because they measure different things. Here is the number read from the network, the window it covers, and why the bigger ones are not comparable.
Solana carried about 1,708 user transactions a second over the 12 hours to 2026-11-16 00:42 UTC, and produced a slot every 214.9 ms against a 200 ms target (both read from the network's performance samples). That is what people actually sent, not a ceiling. Headline figures in the tens of thousands are capacity claims, not counts, and older live counters included validator votes, which were about 73% of all transactions until Alpenglow took votes off-chain on October 21, 2026.
What we measure
- Slot time
- 214.9 ms
- Target
- 200 ms
- User transactions a second
- 1,708
- Busiest minute
- 2,923/s
- Vote transactions
- 0.0%
- Slots skipped, this epoch
- 0.02%
Window: 720 one-minute samples, 12.0 hours to 2026-11-16 00:42 UTC, slots 469,733,991 to 469,934,720. Skipped share: 82 of 350,780 leader slots in epoch 1087 so far.
| Measure | Value | Unit | Read or computed |
|---|---|---|---|
| Mean slot time | 214.9 | ms | computed from samples |
| Median 1-minute slot time | 215.1 | ms | computed from samples |
| Slowest 1-minute slot time | 223 | ms | computed from samples |
| Slot-time target | 200 | ms | protocol (SIMD-0525) |
| User transactions per second, mean | 1,708 | tx/s | computed from samples |
| User transactions per second, busiest minute | 2,923 | tx/s | computed from samples |
| User transactions per second, quietest minute | 921 | tx/s | computed from samples |
| All transactions per second, mean | 1,708 | tx/s | computed from samples |
| Vote transactions in samples | 0 | % | computed from samples |
| Slots skipped, epoch so far | 0.02 | % | read (getBlockProduction) |
| Finality lag | 0 | slots | read (getSlot) |
| Sample window | 12 | hours | read |
Why headline TPS figures are not measurements
They counted votes. Until Alpenglow, every validator sent its votes as ordinary on-chain transactions, and a live "TPS" counter that did not filter them was mostly counting validators talking to each other. We checked old blocks: in 4 blocks from October 9 and 16, 2026, 73% of transactions were votes (from 69% to 78% per block). In a block from November 10, 2026, the share was 0%. Since October 21, 2026 votes travel between validators directly, so the network's own count of all transactions and of non-vote transactions are now the same number. Any TPS figure from before that date, or a chart that spans it, is not comparable with today's.
They are capacity, not traffic. A number in the tens of thousands describes what the software is said to handle under test or in theory. No minute we have read comes close: the busiest single minute in our 12-hour window was 2,923 transactions a second, and the quietest 921. Capacity matters, but it is a different question from how busy the chain is.
They mix in failures. Our count includes transactions that landed and failed, because they still used a slot and paid a fee. In sampled blocks roughly a third of landed transactions fail, most of them automated trading competing for the same opportunity, so "successful user transactions a second" would be lower still. Fees and failure rates are measured on Solana transaction fees.
Block time against the target
A slot is the window in which one validator, the leader, may produce a block. The target was 400 ms for years; four steps between August and October 2026 cut it to 200 ms (dated log). Measured, slots run about 7% slower than the target: 214.9 ms over the window, with the slowest minute at 223 ms. Measured from block times instead, the last 10 epochs averaged 25.8 hours, or 214.7 ms a slot, which agrees.
That gap is why an epoch lasts about 25.8 hours rather than 24 (epoch length, measured).
Finality
Finality is the point after which a transaction cannot be reversed. solana.com gives Alpenglow's target as about 150 milliseconds, down from about 12.8 seconds under Tower BFT (stated, not measured by us). What we can read is how far the network's "finalized" slot trails its newest "processed" slot: 0, 1, 0 slots in three back-to-back readings at 2026-11-16 00:42 UTC. At the old 400 ms slots, 12.8 seconds was 32 slots. A reading of zero or one slot is consistent with near-instant finality, but two separate calls to one node are not a precise stopwatch.
Daily readings
| Date | Epoch | Slot time | User tx a second | Votes | Skipped |
|---|---|---|---|---|---|
| 2026-11-16 | 1087 | 214.9 ms | 1,708 | 0.0% | 0.02% |
| 2026-11-15 | 1086 | 215.1 ms | 2,006 | 0.0% | 0.03% |
| 2026-11-14 | 1085 | 215.4 ms | 2,046 | 0.0% | 0.09% |
| 2026-11-13 | 1085 | 215.2 ms | 2,101 | 0.0% | 0.14% |
| 2026-11-12 | 1084 | 215.0 ms | 2,221 | 0.0% | 0.11% |
| 2026-11-11 | 1083 | 214.9 ms | 1,792 | 0.0% | 0.22% |
| 2026-11-10 | 1083 | 214.2 ms | 2,008 | 0.0% | 0.71% |
How this is measured
Slot time = total sample seconds × 1,000 ÷ total slots over the window. Transactions per second = non-vote transactions ÷ sample seconds (the same as all transactions since Alpenglow moved votes off-chain). Worst and best minute are the 1-minute samples with the slowest slot time and the highest throughput. Skipped share = (leader slots − blocks produced) ÷ leader slots for the current epoch so far. Finality lag = processed slot − finalized slot, median of three back-to-back readings. The pre-Alpenglow vote share is a one-off reading: getBlock (full) on mainnet-beta; transactions with an instruction to Vote111111111111111111111111111111111111111 ÷ all transactions in the block, on 5 blocks, read November 10, 2026.
- Samples come from one RPC node's view of the network and cover at most the last 12 hours; a gap between runs is not seen.
- Throughput is what users sent, not what the network can carry. Most Solana traffic is automated trading, and about a third of landed transactions fail (they still count here).
- The two getSlot calls are not atomic, so finality lag can read one slot high or low.
- Skipped share is for the epoch so far, which can be only a few hours.
Questions people ask
What is Solana's real TPS?
About 1,708 user transactions a second, averaged over the 12 hours to 2026-11-16 00:42 UTC; the busiest minute in that window reached 2,923 a second. Since Alpenglow every transaction counted is a user transaction: votes no longer are.
What is Solana's block time?
214.9 ms on average over the last 12 hours (measured), against a 200 ms target in force since October 9, 2026. Before August 2026 the target was 400 ms.
Why do sites show different Solana TPS numbers?
Three reasons: some count validator vote transactions (until October 21, 2026 these were most transactions), some quote theoretical capacity instead of traffic, and they average over different windows. Ask what was counted and over what period.
How fast is Solana finality now?
solana.com states that Alpenglow targets about 150 ms finality, down from about 12.8 seconds under Tower BFT. In our readings the finalized slot trailed the newest slot by 0 slots (median of three).
Does a higher TPS mean lower fees?
Not directly. Fees depend on demand for the specific accounts a transaction touches. What a transaction costs today is measured on Solana transaction fees.
Sources
SolPilled is independent and unofficial: not the Solana Foundation, Anza, or any validator. Figures are read from public endpoints at the time shown and can be wrong if an endpoint is; for the network's state right now, check status.solana.com and a block explorer. Nothing here is investment advice.