SEPTEMBER 2026 EDITION
Solana Trading Cost Index, September 2026
What one side of a Solana swap cost, all in, on six apps at four order sizes. Collected .
This is a dated edition and its numbers are never edited after publication. If something here turns out to be wrong, the correction goes in a changelog entry rather than a silent rewrite. The current edition is always at the index's permanent URL, and this month's raw data stays online.
All-in cost by platform and order size
Each cell is the median across 20 Solana tokens, in basis points of the amount traded, for one side of a trade. 100 basis points is 1%. Lower is cheaper. Every figure combines the platform's published fee, the price impact measured from a live Jupiter quote, and the network cost described underneath the table.
| Platform | $10 | $100 | $1,000 | $10,000 | Network cost basis |
|---|---|---|---|---|---|
 fomo | 209.9 | 105.8 | 61.0 | 104.2 | covered by platform |
 Jupiter | 13.7 | 11.2 | 11.0 | 54.2 | network median |
 BullX | 2231.6 | 322.9 | 132.2 | 156.3 | platform default |
 Photon | 216.0 | 121.4 | 112.0 | 154.3 | documented floor |
 Axiomnet | 317.0 | 127.0 | 108.1 | 149.4 | platform default |
 GMGN | 120.5 | 111.8 | 111.1 | 154.2 | documented floor |
Axiom carries a net marker because its documentation publishes only rates net of cashback and never states a gross rate, so its column is not strictly comparable with the others. That is a gap in Axiom's documentation, not an estimate we chose to make.
What this edition shows
At the smallest size tested, $10, the spread between platforms is enormous: Jupiter comes in at 13.7 bps and BullX at 2231.6 bps. That gap is not mostly about trading fees. It is the SOL-denominated priority fee and bribe that several terminals default to, which is close to a fixed cost per transaction and therefore swamps everything else on a small order.
At $10,000 the ordering compresses and price impact starts to dominate instead, because impact scales with size while a flat network cost does not. Jupiter is cheapest at that size on 54.2 bps.
fomo is the only platform in this set with size-banded pricing rather than a single percentage, which is why its row does not fall in a straight line as size increases. The bands come from fomo's own help center and are modeled in full on our fee calculator.
Where the cost actually comes from
One token, one size, split into its three parts. GOAT at $100 is shown because it sits mid-basket on liquidity, so neither the deepest nor the thinnest pool distorts the picture. Price impact is identical across every row by construction: execution is held constant so the comparison isolates what each platform charges.
| Platform | Platform fee | Price impact | Network cost | All in | bps |
|---|---|---|---|---|---|
| Jupiter | $0.000 | $0.078 | $0.004 | $0.082 | 8.2 |
| fomo | $0.950 | $0.078 | $0.000 | $1.03 | 102.8 |
| GMGN | $1.00 | $0.078 | $0.011 | $1.09 | 108.8 |
| Photon | $1.00 | $0.078 | $0.106 | $1.18 | 118.4 |
| Axiom | $0.950 | $0.078 | $0.212 | $1.24 | 124.0 |
| BullX | $1.00 | $0.078 | $2.12 | $3.20 | 320.0 |
What this index does not measure
Router quality. Price impact here is measured from a live Jupiter quote and applied identically to every platform. Jupiter is the only one of the six that documents its own router. BullX, Photon, Axiom and GMGN do not publish what performs their swap routing, so a platform routing better than Jupiter would beat these numbers and one routing worse would miss them. We are not going to guess which.
Executed fills. Nothing here is a trade we placed. These are published fee schedules applied to a measured quote. An index built on real fills across six platforms every month would cost real money and one fill is still a single noisy sample.
Whether a schedule is current. None of the five competitor fee pages carries a last-updated date. We can prove only that a schedule was live on the day we read it, which is why the read date is part of the methodology rather than a footnote.
A user-specific priority fee. The network figure comes from a network-wide sample of recent Solana prioritization fees, not from what it costs to land a swap against one specific pool.
Sources for every fee in this edition
| Platform | Published rate | Source, and the date we read it |
|---|---|---|
| fomo | Four Solana size bands, 0.10 USDC flat up to 0.50% | help.fomo.family, “Trading fees on fomo”, read 2026-08-20 |
| Jupiter | Manual Mode (Market Swap): 0%. Ultra Mode: 0-0.5%, tiered by token pair, not order size. | docs.jup.ag/user-docs/trade/spot/fees, read 2026-09-06 |
| BullX | 1% of the initial token amount, collected in SOL for every buy and sell | bullx.gitbook.io/bullx-neo-docs/fees-and-gas, read 2026-09-06 |
| Photon | Photon fees are collected in SOL and it is 1% of the initial token used for every buy and sell transaction. | pies-organization.gitbook.io/photon-trading/photon-on-sol/photon-fees-sol, read 2026-09-06 |
| Axiom | Wood (1X) Net Fee: 0.95% / Cashback 0.05% | docs.axiom.trade/getting-started/fees/axiom-fees, read 2026-09-06 |
| GMGN | GMGN only charges a 1% handling fee for a single transcation, and there are no additional charges. | docs.gmgn.ai/index/gmgn-fees-settings, read 2026-09-06 |
Check it yourself. Every quote behind this edition is published as raw JSON at /data/solana-trading-cost-2026-09.json, including the route Jupiter took for each quote and the slot it was taken at.
SOL was $106.09 when this edition was collected. The network cost column moves with that price, so this index can move month to month even when no platform changes a rate. Median non-zero priority fee in the sample: 36,054 lamports over 150 slots.
What “network cost basis” means in the table above
- platform default
- the platform's own documented default
- documented floor
- a documented minimum, not a typical payment
- network median
- the measured network median, because the platform publishes no figure
- covered by platform
- zero, because the platform says it pays this itself