Crypto liquidations by exchange
Which venue is actually blowing traders up, over the last 24 hours. These are observed forced closes read from each exchange's own public websocket — not an estimate, not a vendor feed, and not a model.
Over the last 24 hours Binance liquidated the most at $78.0M, which is 45.4% of everything our collector saw across nine venues ($171.9M in total).
| Exchange | 24h liquidated | Share | Longs | Shorts | Long / short |
|---|---|---|---|---|---|
| Binance | $78.0M | 45.4% | $52.8M | $25.2M | |
| Hyperliquid | $51.4M | 29.9% | $22.0M | $29.4M | |
| OKX | $18.6M | 10.8% | $13.1M | $5.4M | |
| Bybit | $13.6M | 7.9% | $6.5M | $7.1M | |
| Gate | $8.7M | 5.1% | $1.5M | $7.2M | |
| HTX | $1.3M | 0.8% | $211K | $1.1M | |
| Binance (coin-M) | $280K | 0.2% | $189K | $92K | |
| Bitfinex | $59K | 0% | $59K | $0 | |
| BitMEX | $7K | 0% | $0 | $7K |
Measured by the MarginPad collector across nine exchange websockets · rolling 24h · 2026-08-19 02:20 UTC
Share matters more than the raw dollar figure. A venue carrying more open leverage will always liquidate more money on the same market move, so a big total is mostly a statement about size. The long/short split is the more useful column: it tells you which side the crowd was caught on when the move came. And one day is weather, not climate — a single cascade can hand a quarter of the day's total to a venue that normally sits at five percent.
Why this is hard to find for free
Per-venue liquidation flow is one of the few derivatives datasets that is genuinely gated. The aggregators that publish it charge for the breakdown, because collecting it means holding a live websocket connection to every exchange, normalising nine different message formats, and staying connected through their outages rather than sampling a REST endpoint every few minutes. MarginPad runs that collector for its own liquidation heatmap and live feed, so publishing the per-venue split costs nothing extra and there is no reason to charge for it.
Method, in full
The collector subscribes to the public liquidation stream of nine exchanges: Binance, Bybit, OKX, Hyperliquid, Gate, HTX, dYdX, BitMEX and Bitfinex. Binance is read on two feeds (USD-margined and coin-margined) which are counted as one venue. Each event carries a symbol, a side, a fill price and a size; notional is price times size in US dollars. Events are aggregated into a rolling 24-hour window and the table above is that window at the moment the page was served.
Nothing is inferred. If a venue's socket is quiet it shows zero rather than an interpolated figure, and if the collector itself is down the table says so instead of showing stale numbers as if they were current. Totals will differ from other trackers, and they should: exchanges publish liquidation events at different granularity, some batch them, and any aggregator that models the gaps will report more than one that does not.
Take the data
The same figures are free as JSON with no key and no signup: GET https://marginpad.io/api/v1/venues returns every venue with its 24-hour total, the long and short breakdown and its share. Market-wide totals live at /api/v1/liquidations and the heatmap bands at /api/v1/clusters. Full documentation is on the free crypto API page — attribution is appreciated but not required.
Questions
Which exchange has the most crypto liquidations?
It changes daily, which is why the table above is measured rather than written down. Binance is usually the largest by absolute size because it carries the most open leverage, but share moves with the day: Hyperliquid regularly takes a bigger slice during sharp moves because its book is thinner, and a venue can jump from five percent to a quarter of the day’s total in one cascade. Read the share column, not the raw dollars, when comparing venues of different sizes.
Where does this liquidation data come from?
MarginPad runs its own collector subscribed to the public liquidation websocket of each venue: Binance, Bybit, OKX, Hyperliquid, Gate, HTX, dYdX, BitMEX and Bitfinex. Every forced close is normalised (symbol, side, price, notional, timestamp) and aggregated into a rolling 24-hour window. These are observed events, not estimates or a model, and nothing is filled in when a feed is quiet.
What does the long/short split tell me?
It shows which side the crowd was caught on. A venue whose 24-hour flow is mostly long liquidations was carrying crowded longs into a drop; mostly short liquidations means the squeeze went the other way. The split is often more informative than the total, because it describes positioning rather than size.
Does a bigger liquidation total mean an exchange is riskier?
No. It mostly means more leveraged size is open there. A venue with deep books and many traders will liquidate more dollars than a small one on the same market move. What would indicate risk is a venue liquidating far above its share of open interest, or repeated cascades on thin books.
Can I get this data as an API?
Yes, free and without a key: GET https://marginpad.io/api/v1/venues returns each venue with its 24-hour total, the long and short breakdown and its share of the market-wide figure. The same collector also powers /api/v1/liquidations for market totals and /api/v1/clusters for the heatmap bands.
Related: live liquidation totals · liquidation heatmap · real-time liquidation feed · liquidation statistics by coin