Sonic's homepage sells one number. "Sonic is the fastest settlement layer for digital assets with over 400k TPS," it read when we fetched it at 06:12 UTC today. We measured 8.33 transactions a second.
That is not a snapshot. We read 50,000 consecutive blocks from Sonic's own JSON-RPC, block 80,773,669 to 80,823,668 — twelve hours and 42 minutes of chain, 380,521 transactions, mean block time 0.914 seconds. A second independent node returned byte-identical counts across all 50,000.
And that was the generous window. The token was up 28 per cent over 24 hours when we measured, and DEX volume up 276 per cent. A second 50,000-block window, spanning 30 September and 1 October, gave 2.07 transactions a second.
The claim does not fit the chain
Block gas limit is 5,000,000,000. At the measured block time that is 5.47bn gas a second, and a bare native transfer costs 21,000 gas, so the ceiling for the cheapest possible transaction is about 260,000 a second — below the advertised figure, before anyone does anything useful.
The docs do not walk it back. They state 400,000 TPS and sub-second finality as achieved fact, with no "theoretical" and no "up to" anywhere we could find. They do contradict the homepage: the consensus page says settlement "takes approximately 1-2 seconds", against "Real sub-second confirmation times". Our measurement puts the docs closer to right.
Mean gas used per block was 2,471,979 — 0.049 per cent of the limit. The capacity is not the constraint. Demand is.
Most of the traffic is machines failing
We pulled receipts for 5,768 transactions across 834 blocks. Just over 29 per cent reverted outright. Another 45 per cent either succeeded while emitting no log at all or went to a contract that behaves like an arbitrage searcher. The generous upper bound on genuine economic activity — plain transfers plus logged calls to everything else — is 25.8 per cent, which implies about 2.1 transactions a second of real use.
One unlabelled contract takes 28.7 per cent of all transactions, burns about 350,000 gas a call and succeeds without emitting a log 95 per cent of the time. The top ten destinations take 71 per cent of volume and none resolves to a named public protocol. Our classification is behavioural, not identified, and we would not claim better.
What the money says
Chain TVL is $21.0m, down 98.2 per cent from $1.14bn in May 2025. Stablecoins are $156.9m against a peak of about $604m, so most stable value there is parked rather than deployed. Shadow Exchange takes 44 per cent of 24-hour DEX volume, 42 of it through one concentrated-liquidity venue. Every protocol deployed on Sonic earned $22,256 in fees over 24 hours between them; the chain's own line was $784.
Thirty-seven validators are active. Four hold 37 per cent of the stake and ten hold 70 per cent. The largest, at 12.65 per cent, is 86 per cent self-staked and its funding traces to the foundation's own treasury issuance.
The bull case is not nothing
The engineering works. Across our 100,000 measured blocks the longest pause between blocks was five seconds. A coarser scan of the whole chain found nothing worse than a few minutes, all of it in the first week of December 2024, and the $0.001 fee claim holds — the median fee in our sample was $0.000976.
There was one failure, and a timestamp scan cannot see it. Alchemy's status page records "an accidental hard fork" on 25 March 2026, caused by "a consensus-breaking change introduced in Sonic client v2.1.6", which pushed nodes on older versions out of consensus at block 66,026,364. The chain itself kept producing blocks; the incident ran two days. Sonic's own version history does not mention it.
So: a chain with real headroom, a clean uptime record, a cheap fee and almost nobody using it. That is a better problem than the reverse. It is not a chain doing 400,000 transactions a second.
| Claimed throughput | over 400,000 TPS |
| Measured, 50,000 blocks (12.7h) | 8.33 TPS |
| Measured, 50,000 blocks (30 Sep) | 2.07 TPS |
| Upper bound, genuine activity | 25.8% of tx, about 2.1 TPS |
| Reverted transactions | 29.0% of sample |
| Block gas used, mean | 0.049% of the 5bn limit |
| Chain TVL | $21.0m, −98.2% from peak |
| Protocol fees, 24h | $22,256 |
| S price | £0.0455 ($0.0602) |
| Market cap, equal to FDV | £176.8m ($234.1m) |
| Active validators | 37; four hold a third of stake |
Sources, all retrieved 12 October 2026: Sonic's own JSON-RPC at rpc.soniclabs.com and sonic-rpc.publicnode.com for block timestamps, transaction counts and receipts across two 50,000-block windows; the SFC staking contract for the validator set; soniclabs.com and docs.soniclabs.com for the claim; DefiLlama for TVL, fees, volume and stablecoins; CoinGecko for price, supply and FDV. Contested: throughput depends on the window, so we report both 8.33 and 2.07 TPS; the explorer showed 27.7 as a five-minute spot reading we could not reproduce, and the lifetime average is about 3.83. Our bot-versus-genuine split is behavioural, not identified — about 71 per cent of volume goes to unlabelled contracts — and tying the largest validator to the foundation's treasury is an inference from matching amounts and a shared sender, and shares vary by sample. CoinGecko and the explorer agree that circulating and total supply are both 3,885,497,663 S, so market cap and FDV are the same number, but we could not confirm that figure on chain. Not a recommendation to buy, sell or hold anything.