Solana Trades Had Wider Spreads Outside Aggregators
A September preprint found wider realized round-trip spreads for non-aggregator trades at five of six sampled Solana liquidity venues, using data through Aug. 31, 2026.
A preprint published Sept. 29 found that trades routed outside aggregators had wider realized round-trip spreads than aggregator-routed trades at five of six sampled Solana proprietary automated market makers, or propAMMs. The research covered trading data from Sept. 1, 2025, through Aug. 31, 2026, with later starting dates for newer deployments.
Ozan Solmaz, Lioba Heimbach and Jason Milionis examined propAMMs on Solana, Base and Monad. The Solana analysis covered the six largest measured venues for SOL/USDC: HumidiFi, BisonFi, Tessera, SolFi V2, GoonFi and ZeroFi.
The researchers paired consecutive buys and sells on the same pool that occurred within 25 blockchain slots. The trades had to be within a factor of two in size, include no intervening operator quote update and belong to the same routing category. The researchers then compared aggregator and non-aggregator trades within groups based on pool, day and trade-size range.
Non-aggregator trades had higher overall median spreads at HumidiFi, BisonFi, Tessera, SolFi V2 and GoonFi. The median spreads were 0.86 basis points versus 0.09 basis points for aggregator trades at HumidiFi; 1.01 versus 0.71 at BisonFi; 0.93 versus 0.90 at Tessera; 4.9 versus 3.88 at SolFi V2; and 4.39 versus 0.78 at GoonFi. One basis point equals 0.01 percentage point. ZeroFi recorded the same overall median spread of 1.5 basis points for both categories.
After matching trades within each pool, day and size group, non-aggregator spreads were wider in 68% to 83% of matched cells at the first five venues. The reported exact sign-test p-values were below 0.001. At ZeroFi, the matched non-aggregator median was 0.05 basis points lower.
The measurements represent realized round-trip price gaps, not individual swap fees. The paired fills came from different trades and did not necessarily represent one trader’s completed buy-and-sell transaction. The non-aggregator category could also include flow with different characteristics from aggregator-routed activity.
The study identified stale prices as one possible explanation. A pool can face losses when market conditions change before its quote is updated. PropAMM operators can update prices without waiting for a swap to change the pool’s reserves. The study did not have historical Solana quote data to test whether quoted prices differed from final settlement prices.
The researchers also examined a Tessera implementation on Base and identified six fee components: base, age, priority, mode, default and address. The transaction sender and the contract calling Tessera’s interface could affect default and address treatment, including exemptions and penalties. The finding applies to the inspected Base implementation and does not establish that Tessera’s Solana deployment or other propAMMs use the same rules.
The study documented changes in Base Tessera fees near block boundaries, including lower fees near the end of blocks and higher fees near the start. Those observations concern Base and do not establish the same pattern on Solana.
Wallet fees are separate from the spread measured in the study. Phantom’s swap disclosure, checked Nov. 2, listed a 0.85% fee for most swaps, with exceptions and variable network fees. Jupiter’s documentation listed a 2-basis-point platform fee for SOL-to-stablecoin swaps on its managed order path and no Jupiter swap fee on its customizable build path. Integrator charges and network costs can still apply.
Those disclosures do not identify the route used by a particular wallet transaction. A completed-trade comparison would need to identify the route and venue, account for applicable charges and compare the amount delivered under similar conditions.








