What a crypto swap actually costs

Asked to price the same transfer at the same moment, eight swap providers do not return the same number. Here is the gap between the best and worst quote on twelve pairs, each measured in a single pass so the comparison is like for like:

The spread is widest on stablecoin pairs and narrowest between two major coins, which is the opposite of the intuition that stablecoins are the simple case. On a $10,000 transfer, 1.6% is $160 that depends on nothing but which site you happened to open.

Knowing which provider is best tells you nothing about whether best is good. For that you need the mid-market price — what the asset actually trades for, taken from the public ticker feeds of several large exchanges and compared against the quote.

On USDT to BTC, in one measured pass: mid-market said 0.01191382 BTC. The best provider quoted 0.01181489, which is 0.83% under. The worst quoted 2.44% under. That gap is the margin, and it is the actual price of the service — not a fee line, because nobody shows it as one.

Across the pairs above the best provider sat roughly 0.8% to 1.2% under mid-market. That is the number to carry around: a swap costs about one percent even when you pick well, and two to three percent when you do not.

Sending one transfer through one provider hands that provider both ends of it — the amount, the source, the destination, the timing. Splitting it across several removes that, and it is not free: only one provider can be the cheapest, and a split by definition does not send everything to it.

Measured on a three-way split of 1,000 USDT into BTC — 450 through the best provider, 320 through the second, 230 through the third — the legs returned 0.01177984 BTC against 0.01180497 for the whole amount through the single cheapest. The split cost 0.21%.

That is the honest trade, and a fifth of a percent is a smaller number than most people expect. Whether it is worth paying depends entirely on whether you care that one company holds a complete record of the transfer.

None of it requires trusting this page. Open any two swap services, ask both for the same pair and amount within a minute of each other, and compare. Then look up the pair on any large exchange and compare again — that second comparison is the one nobody invites you to make, and it is the one that shows the cost.

The pair pages here do both automatically and show their working: every provider's quote side by side, the spread, and the distance from mid-market, refreshed through the day. The network names reference covers a different trap — why a wrong chain name comes back as a wrong-coin error.

Every provider was asked for the same pair and amount in a single pass, so quotes are comparable to each other. Mid-market is the median of several large exchanges' public ticker endpoints, never one venue's word — a single source can be stale or thin, and one coin we checked was priced ten percent away from consensus by exactly one exchange.

FixedFloat is absent from the counts above: it returns an authorisation error on this deployment, which is why seven providers quoted rather than eight. Saying so is cheaper than quietly reporting seven as though that were all of them.