DataStore/jupiter-swap-historical-data-solana-swap-events

Jupiter Swap Events on Solana

This dataset contains 1,000 decoded event records from Jupiter, the main swap aggregator on Solana, captured between 09:38:33 and 09:38:39 UTC on 10 June 2026 (slots 425,520,001 to 425,520,017). Each swap event records which AMM filled the trade, the input and output token mints, and the raw amounts on each side.

It's a free sample from datastore.sh. The full Jupiter Swap dataset has 15 tables, delivered as Parquet, with recent or full-history coverage. The same sample is also on Kaggle.

How Jupiter swaps show up on-chain

Jupiter doesn't hold liquidity. It routes each trade through other DEXs such as Raydium, Orca, Meteora and Phoenix, sometimes splitting it across several pools. As the route executes, the Jupiter program logs an event for each leg through a self-CPI call, and those events are the rows in this file.

That has one big consequence for analysis: a single user trade can produce several swap events. A SOL to BONK trade routed SOL → USDC → BONK produces two events, and adding up the input amounts counts the trade twice. Group by sig first to get one row per trade. In this sample, 336 transactions logged one event, 181 logged two, 76 logged three, 16 logged four and 2 logged five.

File

jupiter_swap_cpi_event.csv (1,000 rows, 12 columns)

• ts: block time, UTC
• slot: Solana slot number
• tx_index: position of the transaction within the block
• ix_path: where the event sits in the transaction. [3,1] means inner instruction 1 of top-level instruction 3
• sig: transaction signature (base58)
• r: always 0 in this sample
• program_id: Jupiter v6 program ID, JUP6LkbZbjS1jKKwapdHNy74zcZ3tLUZoi5QNyVTaV4
• decoder_revision: commit hash of the datastore.sh decoder that produced the row
• program: program account passed to the event CPI
• event_authority: Jupiter's event authority account, D8cy77BBepLMngZx6ZukaTff5hCt1HrWyKk3Hnd9oitf
• remaining: extra accounts passed to the CPI, as a JSON array (empty in this sample)
• data: the decoded event as JSON, keyed by event name

What's inside data

• SwapEvent (841 rows): amm, input_mint, input_amount, output_mint, output_amount
• SwapsEvent (158 rows): a swap_events list of 1 to 4 legs bundled into one event, each with input_mint, input_amount, output_mint and output_amount (no amm)
• FeeEvent (1 row): account, mint, amount

amm is the program that filled that leg, not a pool address. Look it up to see which DEX the route went through.

Quick stats

• Time window: 09:38:33 to 09:38:39 UTC, 10 June 2026
• Slots: 425,520,001 to 425,520,017 (14 slots with events)
• Unique transactions: 611
• Unique AMMs routed through (from SwapEvent): 29
• Unique token mints in swap legs: 78

Load it

import pandas as pd, json

df = pd.read_csv("hf://datasets/DataStore/jupiter-swap-historical-data-solana-swap-events/jupiter_swap_cpi_event.csv")

# Split the JSON payload into event name + body
df["event"] = df["data"].apply(lambda s: next(iter(json.loads(s))))
df["body"] = df["data"].apply(lambda s: next(iter(json.loads(s).values())))
print(df["event"].value_counts())

# One row per single-leg swap event
legs = pd.json_normalize(df.loc[df["event"] == "SwapEvent", "body"])
legs["sig"] = df.loc[df["event"] == "SwapEvent", "sig"].values

# How many single-leg events did each trade log?
print(legs.groupby("sig").size().value_counts())

Or with the datasets library:

from datasets import load_dataset

ds = load_dataset("DataStore/jupiter-swap-historical-data-solana-swap-events", split="train")

Before you run numbers

• Amounts are raw integers in token base units. 1 SOL is 1000000000 and 1 USDC is 1000000. Divide by each mint's decimals before comparing tokens.
• Don't sum amounts across legs. Multi-hop routes will double-count, as explained above.
• Account addresses (amm, mints, account) are stored as 32-byte arrays inside the JSON. Convert them to base58 (the base58 package works) to match them with explorers.
• This is a short sample, about six seconds of blocks. Use it to learn the schema, write a parser or test a pipeline. For volume trends, routing studies or backtests, you'll need a longer window.

Things worth trying

Count how many hops routes take and which AMMs show up most as the middle leg. Find the token pairs that never route directly. Work out effective prices per leg and see how much a two-hop route loses between the first input and the final output.

Where can I get full Jupiter swap history?

The Jupiter Swap dataset on datastore.sh ships as partitioned Parquet with typed schemas, manifests and SHA-256 checksums. You choose a recent window or full history and keep the files, with no API quota. The same catalog has Jupiter Perpetuals, Jupiter DCA and Jupiter Limit Order V2, plus the DEXs Jupiter routes through: Raydium, Orca, Meteora and Phoenix. Browse the Solana catalog.

License

MIT

Citation

If you use this data, please credit and link to datastore.sh:

datastore.sh (2026). Jupiter Swap Events on Solana (Sample). https://datastore.sh/datasets/jupiter-swap