Blog · Research · 7 min read
Measured blockchain throughput: September 2026
Across 38 blockchains measured daily in September 2026, Solana's median rate of 3,734 transactions per second led the field, and even with vote transactions removed it exceeded the other 37 chains' median rates added together.
Published · By RampAtlas Research
Key overview
- Category
- Research
- Reading time
- 7 min
- Sections
- 7
- Published
- Solana recorded a median of 3,734 transactions per second across September 2026, or about 1,435 with validator vote transactions removed.
- Even excluding votes, Solana's median rate exceeded the other 37 measured chains' median rates combined, which sum to about 1,046 transactions per second.
- Base processed about 4.4 times as many transactions per second as Ethereum mainnet, at a median of 82.3 against 18.8.
- Bitcoin's median was 9.76 transactions per second, ahead of Avalanche C-Chain, NEAR and Hedera over the same month.
- Twelve of the 38 chains had a median below one transaction per second, including Cardano, Cosmos Hub, ZKsync Era and Linea.
In this post
Across 38 blockchains measured daily in September 2026, Solana's median rate of 3,734 transactions per second led the field, and even with vote transactions removed it exceeded the other 37 chains' median rates added together.
Every figure below was counted by RampAtlas from each chain's own public node, once a day from September 2 to September 30, 2026. None of it comes from a project's own site, and none of it is a theoretical maximum. It is what the networks actually confirmed while we watched. The method is published in full in how we measure chain throughput, and the live leaderboard is at Chains.
38
Chains ranked
measured daily, September 2 to 30, 2026
3,734 TPS
Solana median
about 1,435 TPS excluding vote transactions
18.8 TPS
Ethereum mainnet median
Base, an Ethereum layer 2, ran at 82.3
The full September leaderboard
| Rank | Chain | Median TPS | Daily range | Block time | Samples |
|---|---|---|---|---|---|
| 1 | Solana | 3,734 | 3,272 to 4,528 | 0.31 s | 34 |
| 2 | BNB Smart Chain | 216 | 125 to 315 | 0.45 s | 34 |
| 3 | Aptos | 151 | 129 to 180 | 0.04 s | 30 |
| 4 | TRON | 143 | 53.6 to 191 | 3.00 s | 34 |
| 5 | Sui | 97.6 | 72.7 to 160 | 0.22 s | 34 |
| 6 | Base | 82.3 | 67.4 to 132 | 2.00 s | 34 |
| 7 | Polygon PoS | 63.9 | 46.4 to 101 | 1.51 s | 34 |
| 8 | Monad | 55.3 | 40.5 to 81.3 | 0.31 s | 30 |
| 9 | Stellar | 41.8 | 32.5 to 55.0 | 5.53 s | 31 |
| 10 | TON | 28.8 | 21.9 to 42.4 | 0.40 s | 31 |
| 11 | XRP Ledger | 22.4 | 14.1 to 52.3 | 3.88 s | 34 |
| 12 | Ethereum | 18.8 | 15.2 to 24.7 | 12 s | 33 |
| 13 | Injective | 17.9 | 12.4 to 25.4 | 0.61 s | 30 |
| 14 | OP Mainnet | 17.2 | 12.7 to 37.3 | 2.00 s | 34 |
| 15 | Arbitrum One | 16.8 | 12.7 to 59.6 | 0.24 s | 34 |
| 16 | Celo | 16.4 | 3.43 to 19.8 | 1.00 s | 30 |
| 17 | Algorand | 11.9 | 7.50 to 20.4 | 2.75 s | 30 |
| 18 | Bitcoin | 9.76 | 5.15 to 23.8 | 531 s | 34 |
| 19 | Avalanche C-Chain | 8.41 | 2.82 to 50.8 | 1.04 s | 34 |
| 20 | NEAR | 7.80 | 3.78 to 15.3 | 0.61 s | 31 |
| 21 | Flare | 5.14 | 4.27 to 6.57 | 1.11 s | 30 |
| 22 | Gnosis Chain | 3.45 | 2.42 to 5.09 | 5.10 s | 30 |
| 23 | Hedera | 3.37 | 2.45 to 8.16 | 2.42 s | 31 |
| 24 | Litecoin | 1.47 | 1.26 to 2.59 | 153 s | 26 |
| 25 | Blast | 1.24 | 1.16 to 8.61 | 2.49 s | 30 |
| 26 | Berachain | 1.09 | 0.56 to 1.74 | 2.00 s | 30 |
| 27 | Starknet | 0.96 | 0.47 to 3.19 | 1.65 s | 30 |
| 28 | Sonic | 0.65 | 0.27 to 3.82 | 1.92 s | 30 |
| 29 | Mantle | 0.61 | 0.33 to 0.99 | 2.00 s | 33 |
| 30 | Dogecoin | 0.25 | 0.18 to 0.57 | 62 s | 31 |
| 31 | Osmosis | 0.23 | 0.06 to 1.06 | 1.30 s | 30 |
| 32 | Cardano | 0.21 | 0.10 to 0.33 | 19 s | 26 |
| 33 | Cronos | 0.19 | 0.05 to 0.46 | 0.41 s | 31 |
| 34 | Cosmos Hub | 0.16 | 0.08 to 0.26 | 5.68 s | 30 |
| 35 | ZKsync Era | 0.15 | 0.06 to 0.42 | 7.22 s | 30 |
| 36 | Bitcoin Cash | 0.13 | 0.10 to 0.26 | 568 s | 31 |
| 37 | Linea | 0.12 | 0.09 to 0.25 | 10 s | 30 |
| 38 | Sei | 0.05 | 0.00 to 2.04 | 0.47 s | 30 |
Two tracked chains are not ranked. Hyperliquid has no public node adapter yet, so its throughput has never been measured. Plasma's two public nodes rate-limited our requests on most days, which left only two successful readings in the month, too few to publish a median.
Solana, with and without votes
Solana's headline number needs one qualification, and its own node supplies it. The network counts vote transactions, which are consensus messages that validators send each other, in the same total as user transactions. Its public performance samples report both figures, so we record both.
With votes removed, Solana's median across September was about 1,435 transactions per second. That is still 6.6 times BNB Smart Chain, the next chain on the list. It is also more than the other 37 chains' median rates added together, which sum to about 1,046 transactions per second.
Aptos gets a similar adjustment. It commits two bookkeeping transactions with every block, and its median falls from 151 to about 99 transactions per second without them. That would move it below TRON and Sui in the ranking. The table keeps each chain's full count so the method is identical for every row, and the chain pages print the adjusted figures beside it.
Base carries more than Ethereum itself
Base, a layer 2 that settles to Ethereum, ran at a median of 82.3 transactions per second in September. Ethereum mainnet ran at 18.8. Base alone processed about 4.4 times the transactions of the chain it settles to.
The other large Ethereum rollups were much closer to mainnet. OP Mainnet's median was 17.2 and Arbitrum One's was 16.8. Arbitrum also had the widest spread of the three, with one daily reading at 59.6. ZKsync Era and Linea, two zero-knowledge rollups, sat below one transaction per second for most of the month.
The pattern matches what the layer 1 versus layer 2 guide describes. Activity has moved off Ethereum's base layer, but it has not spread evenly across the layer 2s built on top of it.
Bitcoin and the slow-block chains
Bitcoin's median was 9.76 transactions per second, with daily readings between 5.15 and 23.8. That put it 18th, ahead of 20 chains, including Avalanche C-Chain, NEAR and Hedera, most of which produce blocks far more often.
Chains with slow blocks need a longer window to measure fairly. Bitcoin produced a block about every 531 seconds in our samples, Litecoin about every 153 and Bitcoin Cash about every 568. Their samples therefore span far more than the ten-minute minimum: about two hours of blocks for Bitcoin, about 35 minutes for Litecoin and close to eight hours for Bitcoin Cash.
Twelve chains under one transaction per second
Twelve of the 38 chains had a median below one transaction per second in September. They were Starknet, Sonic, Mantle, Dogecoin, Osmosis, Cardano, Cronos, Cosmos Hub, ZKsync Era, Bitcoin Cash, Linea and Sei.
A low reading is a measurement of use, not of capacity. A chain can be designed for thousands of transactions per second and still confirm a handful when few people are sending anything. The table reports what happened, not what would be possible under load.
Sei's reading is the least stable in the set. Its public node is read over a window of about 23 seconds, so a few quiet blocks can take a daily figure close to zero. Its range of 0.00 to 2.04 reflects that.
How the numbers were collected
A scheduled job reads each chain's own free public node once a day at 06:45 UTC. It walks back from the newest block and counts confirmed transactions over a window of at least ten minutes, or 50 blocks, whichever comes first. The rate is the transaction count divided by the window's length.
Some chains were sampled more than once on a given day because of a manual run, so sample counts range from 26 to 34 per chain. A failed reading is recorded as a failure with the node's reason, never as a zero. Days a chain could not be read are simply absent from its median.
The full method, including how each chain family is handled, is in how we measure chain throughput. Every daily reading behind the table, 1,189 rows with the node's own note for each, is available as a CSV download. Anyone may republish the table or the data with a link to this page.
Frequently asked questions
Why are these figures so much lower than the ones projects publish?
Most published figures are maximums: what a network can process under ideal test conditions. These are realized rates, meaning the transactions a network actually confirmed while we sampled it. A chain rarely runs anywhere near its ceiling, so the two numbers measure different things.
Does a higher rate mean a better blockchain?
No. Throughput measures how much a network is used, not how secure, decentralized or cheap it is. A chain can rank low because few people use it, or because its design favours other properties.
How often is this leaderboard updated?
The measurements run daily, and the live ranking at Chains always shows the latest reading. This post is a dated monthly snapshot, and we plan to publish one for each month.