Tool
Which chain fits the job?
Twenty layer ones on the same fields. Move the dials for what your job needs and the ranking reorders. Kaspa gets no special treatment here, which is the point.
Readings are dated, not live. All twenty rows were re-measured against live endpoints on 22 August 2026.
Start from a job
What matters, and how much
Ranking
How the score works, and where the numbers come from
- Every dial is a weight, not a filter. Turn one up and it takes a bigger share of the score. Nothing is ever removed from the list for failing a requirement.
- Each field is normalized across the chains that report it, then the score is the weighted average of those. The shares shown next to each dial add to 100.
- A chain that never published a field is skipped on that dial rather than scored zero. It is dropped entirely once the fields it does report cover under 70% of what you asked about, and the row says how many dials it stayed quiet on.
- Throughput is what a chain carries now, next to its highest observed burst, for the reason given above. A headline capacity figure is not a measurement.
- Some numbers are not comparable across rows even though they share a column. Nakamoto coefficients use different thresholds for proof of work and proof of stake, developer counts roll up child ecosystems for some chains and not others, and block-producer counts measure different things by design.
- Every row was re-measured on the same day, and the row says which fields were. All twenty were re-measured against live endpoints on 22 August 2026, closing the gap from an earlier pass on 1 August that covered only nine: some readings come from aggregated chain stats, some from timing blocks on the chain's own RPC or mirror node. That pass found real drift rather than rounding: Ethereum's sustained rate ran 18 percent higher than the 1 August reading, Avalanche's ran 62 percent lower, and Internet Computer's came back roughly 3.6x higher on a different metric endpoint than the one the earlier figure likely used. Two fields could not be independently re-derived this pass: Polkadot's throughput stays unscored because the relay chain itself carries almost no user transactions, and TON's throughput carries a same-day correction made earlier rather than a fresh re-derivation. A number nobody re-checked is worth less than one somebody did, so the row tells you which it is.
- "Shows up fast" and "settles for good" are not the same property wearing two names. Across the 14 chains that carry a coded finality rule, the two barely move together (r=0.05, n=14): a chain that includes a transaction quickly is not reliably the one that finalizes it quickly. Kaspa is the clearest case. A block shows up in 0.1s and the coded finality boundary sits at 12 hours. That figure reflects GHOSTDAG. DAGKnight is a research proposal (KIP-2, Status: Proposed) with no mainnet implementation and would only change this if it ships. Bitcoin, Litecoin, Monero and Bitcoin Cash are left unscored on settlement rather than credited with the 6-confirmation convention, so they no longer inflate this comparison either.
- "Pays for its own security" replaced a hash-power dial that ranked Bitcoin's 912.5 EH/s against Kaspa's 0.312 EH/s on one axis. SHA-256, Scrypt, RandomX and kHeavyHash hash rates do not convert to a common unit of work, so that ranking compared five different things as if they were one. Daily issuance in dollars does convert, so the dial now scores that instead: only the five proof-of-work chains here have it sourced (each row shows native units and dollars at spot), and every proof-of-stake, BFT, federated and permissioned chain is left unscored rather than credited with zero. Ethereum is one of those left out on purpose, not by oversight: its net issuance is burn-adjusted under EIP-1559 and moves with fee demand, so there is no single "tokens per day" figure to quote without a netting calculation this pass did not do.
How to read these numbers
Throughput is shown as what a chain carries now and what it has been seen carrying at peak, because those answer different questions. A quiet chain is not a slow one. Kaspa's own breakdown works through why one number never covers it.
Field readings are dated, not live. All twenty rows were re-measured against live endpoints on 22 August 2026, and none of them update themselves.
Check that date against today's before treating any single number as current. Read the whole panel as a snapshot for comparing chains against each other, not as a live feed.
Other demos on this site show fewer chains than this page. Each one lists only the chains it can price on the exact metric it is demonstrating, so the attack-cost and node-cost demos cover five, and the emission-schedule demo covers one.