Pump.fun: 1 in 58 coins graduate, 79% die after day one

01 · Two minutesThe life of a new coin

Open a meme coin app and the list of new coins never stops moving. On a typical day, a fresh one appeared every 2.7 seconds. Each has a name, a picture and a ticker. Most stop trading within minutes.

On 15 June, a dev launched a coin called Le Chaton Fat, French for "the fat kitten", and bought 1.5 SOL of it in the same transaction. The dev is the wallet that creates a coin, and SOL is Solana's own coin. Eight more wallets bought in within five seconds, and the dev sold 14 seconds in. Less than two minutes after the launch, the last of them sold, and nobody has traded the coin since. More than 400 transactions touched it in that time. All but 20 of them failed.

Le Chaton Fat was one of 11.5 million new tokens that traded on Solana in the 11 months to early October. Bitquery records every trade on Solana's DEXs, so we could follow each of those tokens from its first trade to its last. Half of the Pump.fun coins were done within 3 minutes, barely longer than the kitten lasted. 1 in 58 Pump.fun tokens made it to a full trading pool. Hundreds of thousands of others skipped that step. Their devs opened a pool of their own, then emptied it. In the pools we traced, most of the SOL that buyers lost came from wallets tied to the dev.

02 · The machineWhat is a Solana launchpad?

A launchpad is a website that lets anyone create a coin with a few clicks and start trading it at once. Pump.fun is the largest on Solana. Meteora, a Solana exchange, runs one called Dynamic Bonding Curve (DBC), and many apps put their own brand on it. Raydium, another exchange, runs LaunchLab, home of Bonk.fun and, since September, StonkFun.

A new coin starts on a bonding curve, a price formula run by the launchpad. Nobody sits on the other side of a trade. Each buy pushes the price up the curve and each sell pushes it back down, and the SOL that buyers pay stays with the curve.

When enough SOL has come in, the coin graduates, or migrates, as traders also say. On Pump.fun, the SOL and the remaining coins move into a pool on PumpSwap, Pump.fun's exchange, where anyone can trade against them. A pool made by a graduation starts with about 85 SOL, and 98% of these pools had nothing taken out by July. For traders, graduation is the moment a coin gets a real market. Most coins never get there: their curve stops moving when people stop buying.

03 · The countWhich Solana launchpad makes the most tokens?

Pump.fun, by far: it made 78% of all new tokens over the 11 months. The total swung from month to month. September, the busiest full month, had 61% more new tokens than November, the quietest.

Pump.fun's share rose to 90% in August before falling back in September. Meteora's launchpad went the other way: from 23% of new tokens in November to 5% in March. Until August, Raydium's LaunchLab was mostly Bonk.fun: 92 of 100 LaunchLab coins we checked on-chain from those months carried Bonk.fun's settings.

September broke the pattern. LaunchLab launched 153,813 tokens that month, 1.8 times its previous best, in January. Most came from StonkFun, a launchpad that pairs most new coins with other tokens, such as the tokenized stocks NVDAx and SPYx, rather than with SOL: only one in nine of LaunchLab's September tokens was paired with SOL. All 20 LaunchLab coins we checked on-chain from September and October carried StonkFun's settings. Our earlier piece asked whether StonkFun dumps on its holders.

Under the hood, most new tokens now use Token-2022, the newer of Solana's two token standards, which can store a coin's name and picture link inside the token itself. Its share of new tokens went from 20% in November to 92% in August.

04 · The makersWho makes Solana's new tokens?

Pump.fun keeps a record of the wallet that signs each launch, and our copy of those records runs from mid-November to early July. In that time, 1.26 million wallets signed 6.1 million launches. Three in four of those wallets made a single coin.

At the other end, about one wallet in 200 made more than a hundred tokens. Between them, those wallets made 48% of all launches. The busiest made one every 4 minutes from November to July.

A signing wallet is not always a person. Apps that launch coins for their users sign with their own wallets, and launch bots make thousands of tokens per week. The six busiest wallets mostly reused whatever was trending: their most common names were AI, SOL and TRUMP.

05 · The clockHow long do meme coins last?

Not long. Half of all Pump.fun launches had their last trade within about 3 minutes of their first, and a third were finished inside a minute.

To compare launchpads fairly, we asked the same three questions of every coin: did anyone trade it in the week after launch day, in the week starting day 7, and in the week starting day 30? The chart shows the share that saw at least one trade. The table adds a stricter test, ten trades or more.

Four in five Pump.fun tokens, 79%, died after day one: they never traded again after their launch day. In their fifth week, 1 in 65 saw any trade at all, and 1 in 437 saw ten. Meteora DBC coins hang on longer at first, because most graduate at once into a Meteora pool and keep trading there for a while. By the fifth week they are no more alive than Pump.fun's.

06 · The exitWhat is the Pump.fun graduation rate?

By 7 October, 1 in 58 of the tokens launched on Pump.fun in the period had graduated. Most did it on their launch day.

The odds moved a lot. Of the tokens launched in November, 1 in 144 graduated within a month. That rose to 1 in 63 for March launches, fell back from April to June, then jumped: of those launched in August, 1 in 32 graduated within a month. Counted by the month they happened, September had more than four times as many graduations as June.

Graduation also got easier to buy outright. In November, 2% of Pump.fun graduations came after three trades or fewer on the curve, so a few large buys filled it at once. In September, a third did.

Graduation still marks the coins that last. Of graduated Pump.fun tokens, 43% still traded in their fifth week. Of the rest, 1% did.

Pump.fun's own graduation records agree. Counted once per coin, they give 1 in 107 up to July, about what trades give for the same months. The odds over the whole period are better because graduations surged after June.

07 · A formalityWhy do most Meteora launchpad coins graduate?

Meteora's Dynamic Bonding Curve lets each app set its own graduation line, and many set it so low that the first buy clears it. Three in four of the coins that started there graduated, and 35% of all its tokens did so after a single trade. In September, 42% of its new tokens graduated on their first trade.

These graduations mean little. 1.7% of graduated Meteora tokens traded at all in their fifth week, against 2.4% of those that never graduated. A graduated coin lands in a Meteora pool where it can trade for days, which is why the chart in the previous section shows Meteora tokens lasting longer in their first week.

08 · The shortcutHow do PumpSwap rug pulls work?

PumpSwap lets anyone open a pool for any coin, and the wallet that opens a pool can take the money out again. Traders call that a rug pull. In the eight months to 5 July, about one PumpSwap pool in ten came from a graduation. Most of the rest were opened for a coin that had never traded anywhere: 451,402 pools whose coin's first trade came in that pool, within a minute of it opening. We call them curve-skip pools: the coin never touched a launchpad curve.

These pools look like graduations. The median one started with 85 SOL, the same as a pool made by a real Pump.fun graduation. The difference is who can take the money out. A graduation pool almost never loses its SOL. A curve-skip pool belongs to the wallet that opened it, which put in the coin's whole supply, a median of 1 billion coins, and kept the right to withdraw. Holding the whole supply makes that wallet the coin's dev.

99% of these pools were later emptied, almost always by the dev who opened them. The median pool lasted 12 minutes before the dev first took money out, and in 87% of pools that came within an hour.

When it empties the pool, the dev's wallet usually leaves with more SOL than it put in: a median of 40 SOL more. That gain is what the pool's traders lost between them.

Most of the SOL those traders lost, though, came from the dev's side. For 160 of these pools, we traced where each buyer's SOL came from in the days before. 95% of the SOL that buyers left behind came from wallets that had sent SOL to the dev or got SOL from it, or that shared a funder with it. Most of the wallets had no such link, but they lost little: about $1 each on average, or 5 SOL per pool between them.

So the money mostly goes round in a circle. These pools sell the look of a graduated coin with a crowd behind it.

Here is one. On 13 June, a wallet opened a PumpSwap pool for a coin called Manlets with 64 SOL and a billion coins. Over the next 16 minutes, more than a thousand wallets traded it. Then the same wallet withdrew everything, SOL and coins alike, and left with 65% more SOL than it brought. The coin never traded again.

Most of these pools were opened by a different wallet each time: 86% came from a wallet that opened no other. Pool records in our data stop on 5 July. After that we can only count coins whose first trade was in a PumpSwap pool, and from July to September there were about a third as many per month as in February and March.

09 · For tradersWhat should you check before buying a new coin?

Start from the base rate. Four in five tokens from Pump.fun die after day one, and 1 in 437 still has ten trades in its fifth week. Buying a coin in its first minutes is a bet against those odds.

Treat graduation as a weaker signal than it used to be. A third of September's Pump.fun graduations were bought through in three trades or fewer, and on Meteora's launchpad a graduation can take a single trade.

Before buying a coin in a new PumpSwap pool, check how the pool was made. Its first record tells you: Bitquery's Pump.fun API and PumpSwap API return it, along with every trade since. Pools from Pump.fun graduations almost never lose their money. A pool that a dev opened for a brand-new coin can be emptied at any moment, and 99% of them were. A crowd of buyers in such a pool proves little: in the pools we traced, most of the SOL they left behind came from wallets tied to the dev. The token holder API shows who holds the rest of the coins.

10 · MethodHow we did this

We used Bitquery's record of every successful swap on Solana's exchanges in the 11 months to 7 October 2026: the full-history archive to 5 July and Bitquery's live Solana data after that. The same history is sold as files on the Bitquery Data Store, and an AI assistant can query it through the Bitquery MCP server. A token counts as new on the day of its first swap. We left out NFTs and outcome tokens of prediction markets (8 of 8 we checked). We also left out tokens whose first trade was in an older pool, because about half of a sample of those were older tokens trading again. And we left out 426,686 old Pump.fun coins trading for the first time in our data, 22% of them on 16 and 17 December. A fresh Pump.fun curve starts at a fixed price, so a first trade that is a sale, or a buy at any other price, marks an old coin, except on the few days when our archive misses many blocks. Between mid-November and early July, Pump.fun's own launch records settle it: a coin with a launch record is new, and one without is old. On-chain checks agreed with these calls 50 of 53 times for coins we dropped, and 11 of 11 times for coins with no launch record.

The launchpad is the exchange of a coin's first trade. LaunchLab stores the name of the site that launched each coin in its on-chain settings, and we read them for ten random LaunchLab coins from each month. A coin graduates when it starts trading in its launchpad's pool (PumpSwap for Pump.fun, Meteora's pools for Meteora, Raydium's for LaunchLab) and its curve stops trading. Up to 5 July we checked that rule against Pump.fun's own graduation records, counted once per coin: it found all but 23 of 47,539. Of the graduations it found, 3% had no such record, mostly on days when those records have gaps. A public Solana node showed the pool that Pump.fun's graduation creates for 798 of 800 sampled graduations from the whole period. Pump.fun's graduation records also hold repeats: about 3 in 10 come after the coin has already graduated, and half of those sit in the same transaction as a PumpSwap trade. Counted raw, the records put the odds at 1 in 76, so we count each coin once and take graduations from trades. Survival counts any swap, and ten or more swaps, in the week starting on day 1, 7 and 30 after launch day, for coins old enough to have that week in our data.

The curve-skip pools come from our pool records, which run to 5 July. A pool counts when it was not made by a graduation and its coin's first trade came within a minute of the pool opening. The dev's gain equals what the pool's traders lost: we checked a sample of 320 pools trade by trade, and in 256 the two matched to the hundredth of a SOL; in the rest, some of the pool's trades are missing from our records. We checked the scene coin, the Manlets pool and a sample of coins of each kind on a public Solana node: 170 of 180 sampled coins first appeared on-chain less than a day before their first trade. For the funding test, a buyer counts as tied to the dev when SOL moved directly between the two from three days before the pool's day to the day after, or when a wallet that paid the dev in that time also paid the buyer before the pool's day ended. Payments under 0.01 SOL, and wallets that paid a thousand others or more, like exchanges, do not count.

11 · EvidenceAddresses and transactions cited

FAQ

How many coins are created on Pump.fun daily?

On a typical day in the 11 months we studied, about 26,000 new Pump.fun tokens traded for the first time, more on busy days and fewer on quiet ones.

What happens when a Pump.fun coin graduates?

A Pump.fun coin graduates, or migrates, when enough SOL has been paid into its bonding curve. The SOL and the remaining coins then move into a PumpSwap pool, where the coin trades like any other.

What is the Pump.fun graduation rate?

About 1 in 58 of the tokens launched on Pump.fun in the 11 months we studied had graduated by early October. The odds were best for summer launches: of the coins launched in August, 1 in 32 graduated within a month.

Which Solana launchpad makes the most tokens?

Pump.fun. It made 78% of the new tokens that traded on Solana in the 11 months we studied. Meteora's DBC made 10% and Raydium's LaunchLab 4%.

How long do meme coins last on Pump.fun?

Half stop trading within about 3 minutes of their first trade, and four in five never trade after their first day. Only 1 in 437 still has ten trades in its fifth week.

What is a bonding curve on Pump.fun?

A bonding curve is a price formula that a launchpad uses to sell a new coin. Each buy raises the price and each sell lowers it, so the coin can trade from its first second without anyone providing a market.

What is a Pump.fun rug pull?

The rug pulls we counted happen on PumpSwap, Pump.fun's exchange, where anyone can open a pool for a coin and keep the right to withdraw its money. In a rug pull, the dev withdraws the pool's SOL, including what buyers paid, and leaves the coin with no market. In the pools we traced, most of the SOL that buyers lost came from wallets tied to the dev, and outside wallets lost about $1 each on average. Pools made by Pump.fun graduations almost never lose their money this way.