THORChain vs Chainflip vs NEAR Intents
We route live volume across these rails every day. Here is how they differ in architecture, fees, speed and what happens when things go wrong.
THORChain, Chainflip and NEAR Intents solve the same problem — swapping native coins across chains — with three genuinely different architectures. We route production volume over all three (and Maya, THORChain's cousin), so this comparison is written from operating data, not whitepapers.
Short version: they differ most in how prices are formed and what happens at the edges — big trades, thin pairs, failures. For a typical mid-size swap on a liquid pair, all three deliver within a fraction of a percent of each other, which is exactly why routing across them beats loyalty to any one.
Three ways to settle the same swap
THORChain is the vault model: ~100 validators bonded in RUNE collectively control vaults on every connected chain, and prices come from continuous liquidity pools (asset ⇄ RUNE). It's the oldest design, with the deepest history on BTC pairs and the most battle scars — it has halted, recovered, and kept user funds whole through several incidents.
Chainflip replaces pools with a just-in-time auction: 150 validators run MPC vaults, and when your deposit lands, market makers bid to fill it against the protocol's virtual AMM. JIT pricing tracks CEX prices closely on majors, and every swap carries an on-chain minimum-price floor — slip past it and the deposit refunds.
NEAR Intents drops pools entirely: you sign an intent ("this deposit for at least X of that asset"), and competing market makers race to fulfil it from inventory they hold on both chains. Pricing is whatever the most aggressive maker will do — often the best quote on the board — and coverage extends to chains no pool protocol supports, because a maker only needs an ordinary wallet on each chain.
Side-by-side comparison
| THORChain | Chainflip | NEAR Intents | |
|---|---|---|---|
| Price formation | On-chain liquidity pools | JIT market-maker auction | Market-maker competition |
| Coin coverage | ~10 chains, majors | 6 chains, majors + stables | 25+ chains, 190+ assets |
| Typical speed | 5–20 min | 2–6 min after confirmation | 1–5 min after confirmation |
| Large trades | Deep on BTC/ETH; price impact grows smoothly | Best near CEX price on majors | Depends on maker inventory per pair |
| Failure mode | Refund minus outbound fee | On-chain min-price floor; auto refund | Quote deadline; refund to your address |
| Memo required | Yes (encodes the swap) | No — channel per swap | No — address per swap |
Fees: same shape, different names
All three charge the same three things economically: outbound network fees, a liquidity/maker margin, and a protocol take. THORChain's cost lives in pool slip plus a ~0.02 RUNE outbound fee; Chainflip's in auction spread plus fixed network fees; Intents' entirely in the maker's quoted spread. On a $1,000 major-pair swap the delivered difference between them is usually under 0.3% — smaller than the movement between two refresh clicks.
Practical consequence: quote all rails at the moment of the swap and take the best delivered amount. That is literally what a router does — Moss quotes every rail carrying your pair and shows the winner, with the runners-up listed for transparency.
When each rail wins
- BTC-heavy, size-heavy: THORChain's deep BTC pools absorb large orders predictably.
- Tight pricing on majors: Chainflip's JIT auction frequently prints closest to spot for ETH/BTC/SOL/USDC.
- Long-tail coverage: NEAR Intents reaches XRP, DOGE, TON, ZEC, TRON and other chains pool protocols don't touch — and often wins majors outright.
- No single rail wins even a majority of quotes we serve — the winner flips with inventory, gas and pool depth throughout the day.
Which is safest?
Different trust profiles: THORChain has the longest incident-tested history; Chainflip enforces price floors on-chain; Intents minimizes protocol surface by keeping funds with makers until atomic settlement. All three settle without custodial accounts, and all three have processed real failures into refunds rather than losses.
Why does Moss route across all of them instead of picking one?
Because the best rail changes per pair, per amount, per minute. Routing captures each protocol's strength — pool depth, auction pricing, maker coverage — and turns rail failures into a re-route instead of an error page.
What about Maya Protocol?
Maya is a THORChain fork sharing the vault-and-pool design with its own asset set (notably native DASH and KUJI). Everything said about THORChain's model applies; depth is smaller.