Hyperliquid datasets
Free Parquet files of Hyperliquid data, built from our own nodes, Hyperliquid's archive and its API. On Hugging Face and Kaggle, under CC BY 4.0.
Daily sets get a file for each UTC day. Monthly sets are refreshed after each month closes. For anything newer, the API serves the same data up to the latest block.
| Dataset | Coverage | Rows | Size | Updates |
|---|---|---|---|---|
| Hyperliquid liquidations | 25 May 2025 to 31 Oct 2026 | 2.85 million liquidations in 8.06 million fills, 468 markets | 290 MB | Monthly |
| Hyperliquid trades | 3 Oct 2026 to 6 Nov 2026 | 35 days, 203,836,201 trades | 5.1 GB | Daily |
| Hyperliquid funding rates | 12 May 2023 to 31 Oct 2026 | 6.1 million rates, 494 markets | 74 MB | Monthly |
| Hyperliquid hourly candles | 25 May 2025 to 31 Oct 2026 | about 3.6 million candles, 799 markets | 129 MB | Monthly |
| Hyperliquid open interest | 20 May 2023 to 31 Oct 2026 | 5.94 million rows, 553 perps | 140 MB | Monthly |
| Hyperliquid HIP-4 outcome markets | 1 Aug 2026 to 6 Nov 2026 | 98 days, 14,314,812 fills, 40,062 contracts | 551 MB | Daily |
The datasets
Hyperliquid liquidations
Every liquidation on Hyperliquid perps, native and HIP-3, in two tables: one row per liquidated position (events) and one per liquidation fill (fills). One Parquet file per month.
- Coverage
- 25 May 2025 to 31 Oct 2026
- Rows
- 2.85 million liquidations in 8.06 million fills, 468 markets
- Size
- 290 MB in 36 Parquet files
- Updates
- Monthly, after each UTC month closes
- License
- CC BY 4.0
- Name
ironflow/hyperliquid-liquidations
Hyperliquid trades
One row per trade on Hyperliquid perps and spot, with the buyer, the seller and the taker side. One Parquet file per UTC day.
- Coverage
- 3 Oct 2026 to 6 Nov 2026
- Rows
- 35 days, 203,836,201 trades
- Size
- 5.1 GB in 35 Parquet files
- Updates
- Daily: a file for the previous UTC day
- License
- CC BY 4.0
- Name
ironflow/hyperliquid-trades
The Kaggle copy is updated less often. Hugging Face has the newest file.
Hyperliquid funding rates
Every settled funding rate on Hyperliquid perps, native and HIP-3, with the premium, from Hyperliquid's fundingHistory. Hourly since 8 June 2023, 8-hourly before. One file per month.
- Coverage
- 12 May 2023 to 31 Oct 2026
- Rows
- 6.1 million rates, 494 markets
- Size
- 74 MB in 42 Parquet files
- Updates
- Monthly, after each UTC month closes
- License
- CC BY 4.0
- Name
ironflow/hyperliquid-funding
Hyperliquid hourly candles
Hourly OHLCV for every Hyperliquid perp and spot pair, built from every fill, so delisted and HIP-3 markets are in it. One file per month.
- Coverage
- 25 May 2025 to 31 Oct 2026
- Rows
- about 3.6 million candles, 799 markets
- Size
- 129 MB in 18 Parquet files
- Updates
- Monthly, after each UTC month closes
- License
- CC BY 4.0
- Name
ironflow/hyperliquid-candles-1h
Hyperliquid open interest
Hourly open interest, mark price and oracle price for every Hyperliquid perp: Hyperliquid's archive until 2 August 2026, Ironflow's own poller since. One file per month.
- Coverage
- 20 May 2023 to 31 Oct 2026
- Rows
- 5.94 million rows, 553 perps
- Size
- 140 MB in 42 Parquet files
- Updates
- Monthly, after each UTC month closes
- License
- CC BY 4.0
- Name
ironflow/hyperliquid-open-interest
Hyperliquid HIP-4 outcome markets
Every fill on HIP-4 outcome markets, one row per fill and one file per UTC day, plus a list of every contract with its event, outcome and settlement.
- Coverage
- 1 Aug 2026 to 6 Nov 2026
- Rows
- 98 days, 14,314,812 fills, 40,062 contracts
- Size
- 551 MB in 99 Parquet files
- Updates
- Daily: a file for the previous UTC day
- License
- CC BY 4.0
- Name
ironflow/hyperliquid-outcomes
Load one in Python
pandas reads a file straight from Hugging Face with huggingface_hub installed:
import pandas as pd
# one month of liquidated positions, read straight from Hugging Face
events = pd.read_parquet("hf://datasets/ironflow/hyperliquid-liquidations/data/events/2026-10.parquet")
print(events.nlargest(5, "notional_usd")[["timestamp", "market", "position_side", "notional_usd"]])Or load a whole set with the datasets library. The liquidations set has two configs, events (the default) and fills; the outcomes set has fills and markets.
from datasets import load_dataset
funding = load_dataset("ironflow/hyperliquid-funding", split="train").to_pandas()
btc = funding[funding.market == "BTC-PERP"].set_index("timestamp")
print(btc.funding_rate.resample("D").sum().tail()) # BTC funding per dayEach dataset card on Hugging Face lists every column and the checks we ran against Hyperliquid's own API.
Newer data
The API serves every fill since 25 May 2025, liquidations, funding, candles and open interest up to the latest block, over REST, SQL and MCP. Without a key it reaches 24 hours back; a free key reaches 30 days, and paid plans up to 1 year.
- Hyperliquid historical data: what Hyperliquid's API and archive keep, and every way to get the full history.
- Hyperliquid SQL: query every fill, liquidation and transfer with SQL.
- Live liquidations: every liquidation as it happens.
License
Every dataset is CC BY 4.0: use it for anything, commercial work included, and credit “Ironflow (ironflow.sh)”.
Wallet addresses come from public Hyperliquid data. To ask us to leave one out of future versions, write to [email protected].