ironflow/hyperliquid-funding

Hyperliquid Funding Rates

Every funding payment on Hyperliquid perpetuals from 12 May 2023 to 31 October 2026: 6,100,462 settled rates across 494 markets, 235 native perps and 259 HIP-3 builder markets (stocks, commodities, indices and pre-IPO names on xyz, para, hyna, km, cash, flx, vntl, io and mkts).

Since July 2023, BTC longs paid shorts about 14.2% a year on average, ETH 13.9% and HYPE 20.2%. In 42.6% of market-hours the rate sits exactly at Hyperliquid's baseline of 0.00125% an hour. Funding hit the 4% an hour cap 18 times, every one on a HIP-3 market (vntl:SPACEX, vntl:OPENAI, vntl:ANTHROPIC, km:USBOND, hyna:ENA); the most negative hour is CYBER at -2.94% on 2 September 2023.

These are the settled rates from Hyperliquid's own fundingHistory, not the predicted rate shown before settlement. The same data is public on Hyperliquid's API, 500 rows per request per market; this is the whole history in one download, collected by Ironflow market by market and checked row for row against the API.

Get live data

Live Hyperliquid data is on Ironflow. The same rates are on Ironflow's API: GET /v1/funding. No key is needed for the last 24 hours; a free key gives 30 days.

Source and updates

• Source: Hyperliquid's public fundingHistory API, collected by Ironflow market by market; settled rates only.
• Checks: sampled rows match fundingHistory exactly; no duplicate (market, timestamp) rows (see Notes).
• Updates: refreshed monthly, after each UTC month closes. For anything newer, use the API.
• Coverage: 12 May 2023 to 31 October 2026. Native perps from the start, HIP-3 markets from 13 October 2025.

Files

One Parquet file per month (UTC): data/YYYY-MM.parquet, one row per market per funding payment.

Column	Type	Meaning
timestamp	timestamp (ms, UTC)	settlement time as Hyperliquid reports it (a few ms after the hour)
market	string	Ironflow market symbol, e.g. BTC-PERP, xyz:NVDA-PERP
coin	string	Hyperliquid's name for the market, e.g. BTC, xyz:NVDA
dex	string	HIP-3 builder dex (xyz, para, ...); empty for native perps
funding_rate	float	the rate applied at this payment: positive means longs pay shorts
premium	float	the premium component Hyperliquid reports with the rate; null in 4,427 rows (see Notes)

Use

import pandas as pd

f = pd.read_parquet("hf://datasets/ironflow/hyperliquid-funding/data/")
btc = f[f.coin == "BTC"].set_index("timestamp").funding_rate
print((btc.resample("D").sum() * 365 * 100).tail())  # daily funding, annualised %

-- DuckDB: markets with the highest average funding over the last 30 days
SELECT market, avg(funding_rate) * 24 * 365 * 100 AS apr_pct, count(*) AS hours
FROM 'data/*.parquet'
WHERE timestamp >= TIMESTAMP '2026-10-01'
GROUP BY market HAVING hours > 600
ORDER BY apr_pct DESC
LIMIT 10;

Notes

• Hyperliquid paid funding every 8 hours until 8 June 2023 and every hour since (BTC's last 8-hour payment is at 00:00 UTC, the first hourly one at 01:00). Rows before that date are 8 hours apart; annualise by the actual spacing, not by 24 a day.
• One hour is missing for every market, 13:00 UTC on 15 August 2024: Hyperliquid's own history has no payment there.
• premium is null where Hyperliquid returns NaN: 9 markets between 15 March and 1 July 2024, around their delisting, with a funding rate of 0.
• HIP-3 markets start on 13 October 2025. Delisted markets are kept up to their last payment.
• 8 randomly sampled rows across the whole range match Hyperliquid's fundingHistory exactly; there are no duplicate (market, timestamp) rows.

Changes

• v1, 1 November 2026. First release.

License and citation

CC BY 4.0. Credit "Ironflow (ironflow.sh)" when you use or share the data.