The League

What Solana trading bots actually charge, read from on-chain fee payments — trade by trade, not from their marketing pages.

Fee as a share of your trade

On the Trojan line: it is drawn as if the ~1% applies to every trade. We measured that percentage stacking on roughly 70% of Trojan trades, not all of them, so Trojan’s real curve sits below the line shown here. We draw the conservative version until we can publish a firmer figure.

Both axes are logarithmic, so a flat fee falls as a straight line and a percentage fee is horizontal. Photon is brutal on small trades and unbeatable on large ones; the lines cross around 0.35 SOL.

Who trades where

Every fee payment we read carries the size of the trade that produced it. Axiom is measured from its own fee accounts separately, because it executes through a private engine our main census excludes. Group those by bot and you can see what kind of trader each one attracts. These are shares, not volumes — we sample a fixed number of transactions per fee account, so the number of trades we saw says nothing about a bot's market share. Only the mix, the median and the whale share compare across bots.

Named bots

Cost by trade size (SOL, fees only)

Every fee account we measured

Live accounts collecting fees from Solana traders, ranked by how many separate wallets paid them. Click any column to sort. A name appears only where the bot's own program signed the transaction — everything else stays unnamed, because guessing is how fee comparisons go wrong. Rates are lower bounds: if a bot splits its fee across several accounts, its real cost is higher.

Coverage

The part nobody tells small traders

What about speed?

What you actually get filled at

Fees are only half of what a trade costs. The other half is the price you get, and nobody publishes it. So we measured it: 17,121 real fills across 264 tokens, comparing every trade against the volume-weighted price of other trades on the same token, in the same 30 second window, at a similar size. Same token, same moment, same size, different bot.

Negative is better: the trade got a better price than others happening at the same moment. GMGN is the only bot with a clear edge — about 75 basis points, consistent across two deliberately different sets of tokens. Photon's range crosses zero, so on this evidence it has no measurable edge in either direction. Axiom's two samples disagree by 147 basis points, so treat its number as unsettled.

Total cost: fee plus fill

Method and limits: execution is measured against other traders in the same window, not against a theoretical best price, so it captures being early or late as well as routing quality. It cannot measure sub-second speed claims. It does not control for who chooses which bot or why. All tokens are Solana memecoins in their first hours of trading.

How this was measured

Every number comes from the blockchain, not from a fee schedule. We read the SOL that left traders' wallets on each trade and grouped it by which account received it. Before comparing anything we sorted those accounts into what they really are: a bot's own fee account, a Jito tip, a venue protocol fee, a private execution engine, or routing infrastructure. All eight of Jito's tip addresses sat in our raw data and would have been counted as a trading bot had we not separated them — most fee comparisons never make that distinction. Accounts marked as sitting inside a private execution engine are real fees paid by real traders, but that engine is shared, so we do not attribute them to a particular bot.

Not measured here: speed, fill quality, MEV protection, reliability. Price only. Fees change without notice, which is why the measurement date is printed at the top.

Disclosure