Operator-controlled propAMMs executed quiet-market Solana (SOL) swaps at a reference-relative cost of 0.26 basis points, roughly one-tenth of the 2.59 basis points recorded for public automated market makers, according to a Sept. 29 preprint authored by researchers listing ETH Zurich and Category Labs affiliations.

The study, published on arXiv, covers Sept. 1, 2025 through Aug. 31, 2026 on Solana, with shorter comparison samples on Base and Monad. It weights fills by notional against Bybit's size-weighted top-of-book USDT microprice, converted through the exchange's USDC/USDT midpoint.

What does the pricing gap mean for traders and depositors?

The paper reports two-second gross maker markouts of +0.37 basis points for propAMMs versus −0.22 basis points for public AMMs across the Solana sample. A markout compares a fill against a later reference price; a positive number favors the maker.

The two measures answer different questions. Execution-cost proxies ask what a trader gives up against a stable reference, requiring less than 1 basis point of reference movement from five seconds before to one second after a fill. Markouts ask what happens to a trade's value after the pool accepts it.

The authors are explicit that gross markouts do not establish net liquidity-provider returns. A depositor's economics also reflect asset exposure, fee allocation, inventory changes, hedging, operating expenses and transaction costs. The paper's short horizon leaves that accounting unresolved, and venue averages cannot show that professional pools caused aggregate passive-LP losses.

Low execution cost can attract swappers without being a sufficient investment case for depositors, who supply the inventory others trade against and need compensation for the risks that inventory carries.

How do operators defend their quotes?

Jump Crypto described the propAMM model, including its own BisonFi, in an April account: pools adapt prices and available liquidity to inventory, quote freshness, and the quality of incoming flow. A maker holding too much of one asset can discourage trades that add more of it; a stale price can justify withdrawing depth or widening a fee.

Jump is an interested operator, and implementations differ. Economically, those controls can let a maker quote more tightly when it expects less risk. Requiring identical terms for every counterparty would remove one way of distinguishing that risk, though the price an ordinary swapper ultimately receives still needs to be measured.

Jupiter's AMM integration documentation shows a dedicated signer identifies trades originating at its frontend, describing that flow as retail and non-toxic. Identifying origin, however, is not the same as independently establishing that every trade is harmless to the maker.

What did the Base sample show?

For Tessera on Base, execution averaged 1.08 basis points worse by trade count and 0.56 basis points worse by volume than reconstructed previous-block-end quotes. The researchers label the block-timed fee pattern "spoofing."

The measurement compares reconstructed pool output with execution; individual screen quotes sit outside the measure, and the observed pattern provides no direct evidence of operator intent. Better execution against a market reference and worse execution against an earlier quote can coexist.

In a March 20 report, routing provider 0x described Base prices deteriorating between quote selection and settlement through block timing and spread changes, and stated a policy of cutting off sources until execution issues are remedied. Its operators went unnamed, so that report cannot identify Tessera as the subject.

Can routing close the display-to-fill gap?

Jump argued that routers selecting executable prices when transactions run can largely close that gap. The design implication: a maker could retain inventory, freshness and counterparty protections, provided the router compares outputs that already include them.

Jupiter's current Swap API documentation describes competition between routing engines and a mechanism that sidelines underperforming sources, and its integration guide requires quote/execution parity tests against the same pool snapshot. A parity check measures agreement on one snapshot, though persistence through later updates is a separate question.

A meaningful comparison would require executable output reflecting the same trade size, caller, current pool state and applicable charges. Jupiter documents a platform swap fee on its Meta-Aggregator path and none on its Router path, while integrator fees and landing arrangements differ; a protocol-level spread cannot stand in for the amount ultimately received.

The next useful evidence, the paper's framing suggests, would compare quoted and delivered output on matched transactions and show how routing treats persistently underperforming sources. For passive liquidity, a separate position-level return assessment remains outstanding.