Best Solana sniper bot: how to compare them honestly

Speed is the wrong first question

Almost every comparison of Solana sniper bots opens with latency, and almost every one of them is measuring something the buyer cannot verify. Nobody publishes the node they run on, and a millisecond claim in marketing copy isn't a number you can test before paying.

Speed also matters less than it sounds. Solana targets a 400 millisecond slot, and once two bots both see a mint inside that window, what separates them is the compute unit price they bid and whether they route publicly or through a bundle. Both of those are settings you control, so they aren't really a property of the software at all.

The differences that show up in your balance after a month are duller than latency and much easier to check. Work through the seven below before you pay anyone.

One: where the signing key lives

This is the only criterion where a wrong answer can cost you everything rather than some of it, so it belongs first.

A custodial balance is the model to be most careful with. Depositing SOL into a bot's internal balance means the operator can move it, and the fact that most of them never do isn't a control you hold. If a bot asks you to fund an account rather than connect a wallet, that answer alone should decide whether you continue.

Our own answer sits in the middle and we would rather say so plainly than dress it up. The engine holds a key it can sign with, sealed with AES-256-GCM before storage and never returned to any browser, and your SOL never leaves your own wallet. It's still a key on someone else's machine, so the wallet you link should hold what you're willing to trade and nothing more.

Two: what gets screened before a buy

Ask any bot which checks run before it spends money. The answers cluster into three groups, and the gap between them is wider than any speed difference you will ever measure.

• Two checks. Mint authority and freeze authority, read straight off the mint account. Cheap to implement, and on a Pump.fun launch both are revoked by the program anyway, so on the busiest venue on Solana they tell you nothing at all.
• Four or five checks. The two above plus an LP burn read and sometimes a sell simulation. This is where most paid bots sit, and the sell simulation is the single most valuable one in the group because it catches honeypots at zero cost.
• Ten or more checks. Holder concentration, deployer history, funding graph clustering, Token-2022 extensions and LP lock depth on top. Slower to build, far harder to fake in marketing copy, and the reason a log fills with skips rather than fills.

There is a test that cuts through the claims. Ask to see what a skipped launch looks like. A bot that screens properly can show you a line saying which check refused a specific mint and why, because it had to record that to make the decision. A bot that lists checks it doesn't really run has nothing to show, and the answer turns into an explanation of why logs aren't useful.

Our screening runs six checks on the 1 SOL plan, ten on the 3 SOL plan and twelve on the 5 SOL plan. Every skip is written into the console with the reason attached, and each fill links to its signature on Solscan so the chain can contradict us if we are wrong.

Three: flat or percentage, and the maths nobody publishes

Percentage bots usually charge somewhere around 1% of each trade, which means roughly 2% of a round trip once you count the buy and the sell. Flat bots charge a monthly figure and take nothing from a fill. Which is cheaper depends entirely on how much you trade, and the crossover is easy to work out.

Multiply your buy size by the number of round trips you make in a month. That number is the volume the percentage applies to. At 1% a side, a flat plan wins once your monthly entry volume passes fifty times the plan price.

So if you trade twice a week at 0.1 SOL, a percentage bot is genuinely cheaper, and we'd rather you knew that now than found out later. Flat pricing starts winning around the point where sniping stops being an experiment, and by the time you are running 0.5 SOL entries several times a day the difference is not close.

There is a second effect that no table captures. A percentage of every fill quietly rewards the operator for making the bot trade more, and the settings that make a bot trade more are exactly the settings that lose money. That incentive is worth more of your attention than the arithmetic above, because it doesn't show up on any pricing page and it shapes every default the bot ships with.

Four: does it run while you sleep

Browser tabs throttle timers, drop websockets and suspend entirely when a laptop closes. Anything that has to react inside a 400 millisecond slot cannot depend on a tab staying awake, which rules out extensions for serious use no matter how good the interface is.

That leaves two workable shapes. Either the bot runs on a server the operator maintains, or it runs on one you maintain. Both are fine. What isn't fine is a bot that claims to be always on while quietly living in your browser, and the way to tell is simple: close everything, wait an hour, then check whether anything happened.

Ask one more question while you are there. When the bot is offline, what happens to positions that are already open? An engine that stops watching your exits when it restarts is worse than no automation, because you thought a stop loss was armed and it wasn't.

Five: what all DEXes usually means

Nearly every bot claims full Solana coverage. In practice that phrase covers three quite different things, and only one of them is what you want.

Routing through an aggregator isn't detection. An aggregator answers the question of where to fill an order you already decided to place, and a sniper's whole job happens before that. So the question to ask isn't how many venues it can trade on. It's how many launchpad programs it decodes.

We cover 42 venues and split them honestly: 10 launchpads decoded individually, 26 AMMs, 3 order books and 3 routers. How many you can arm at once depends on the plan, which is a limit we set rather than a limit of the software.

Six: the exit rules nobody compares

Entry gets all the marketing and exits decide the result. A bot that fills beautifully and then leaves you managing the position by hand has automated the easy half.

Five questions separate a real exit engine from a take profit field:

1. 1When are the exits armedAt the moment the buy confirms, or the next time some loop gets around to it? A gap of even thirty seconds is where a rug happens.
2. 2Is there a trailing stopA fixed take profit caps your best trades at the number you guessed. Trailing lets the rare runner run and still protects most of the gain.
3. 3Is there a time capPositions that go nowhere tie up your concurrency limit and your attention. A maximum hold closes them without an argument.
4. 4Does it react to liquidity leavingPrice based exits are always late to a liquidity pull. Watching the pool itself is the only exit that fires on the cause instead of the symptom.
5. 5Is there a daily stopOne bad hour with no daily loss limit can undo a good month, and no entry setting protects you from that.

If a bot answers no to the fourth question, it doesn't really have rug protection, whatever the landing page says. Everything else on that list is table stakes.

Seven: how much reasoning it shows you

A sniper bot spends most of its time refusing things. If you cannot see the refusals, you cannot tell the difference between good screening and a bot that is simply not running.

The log is where that shows. A useful one names the mint, names the check that refused it, and links every fill to a signature you can open on an explorer. A useless one shows fills only, which flatters the bot and tells you nothing about the ninety percent of its work you're paying for.

There is a related question worth asking before you pay: how long is history kept? A log that resets every session can't answer the one question that matters in week three: are your rules working, or have you just had a good few days?

The checklist, in the order that matters

Run any bot through these nine questions before you pay for it. Anything that can't answer the first three clearly isn't worth reaching the fourth.

• Do I deposit funds, or does it sign from a wallet I control?
• How is the signing key stored, and can it ever be read back out?
• Which checks run before a buy, and can you show me a skipped launch with its reason?
• What does it cost at my buy size and my trade count, not at the example on the pricing page?
• Does it keep running with my laptop closed, and what happens to open positions if it restarts?
• How many launchpad programs does it decode, as opposed to how many venues it can route through?
• Are exits armed with the entry, and is there a trailing stop, a time cap and a liquidity trigger?
• Is there a daily loss limit that stops the engine without me being there?
• How long is the log kept, and does each fill link to a signature I can verify myself?

Our answers, in the same order: no deposit; AES-256-GCM and never returned to the browser; six to twelve depending on plan, with reasons in the console; 1, 3 or 5 SOL a month with nothing taken from a fill; yes, and open positions keep their exits; ten launchpads decoded individually out of 42 venues; yes to all four; yes; and 7, 90 or 365 days with a Solscan link on every fill.

Where we lose against some alternatives is worth naming too. A self-hosted script keeps the key on your own machine, and no hosted engine can match that. A percentage bot is cheaper if you trade small and rarely. If either of those describes you, we're not the right tool and there's no point pretending otherwise.

Questions people ask

These come up in the ticket queue often enough to answer once, properly. If yours is not here, send it over and it probably ends up on this page.

Key takeaways

• Latency is the least verifiable claim in this market and the least important once both bots see the mint inside one slot.
• Custody comes first: a bot holding your SOL in an internal balance is a different risk class from one that signs for a wallet you own.
• A percentage fee is cheaper for small, infrequent trading; flat pricing wins past roughly fifty times the plan price in monthly entry volume.
• Ask to see a skipped launch with the reason attached. Bots that only screen in marketing copy cannot produce one.
• Exits decide results, so check that they are armed with the entry and that something reacts to liquidity leaving.
• A self-hosted bot keeps the key on your own machine, and no hosted engine can match that.