Liquidation is not a single event where "the exchange closes your position at your liquidation price." It is a pipeline with several stages, and where a venue sits on each stage determines how much of your margin you actually lose. Here is the sequence, and the parts that genuinely differ.
Your position is marked continuously against a mark price (an index of spot prices, not the last traded price on that venue — this prevents a single wick from liquidating you). When your equity falls to the position's maintenance margin requirement, liquidation triggers. That requirement is tiered by size: on BTC, tier-1 maintenance margin is 0.40% on Bitget and OKX and 0.50% on Bybit, rising each tier as the position grows (full tier tables). The mark-price detail matters: your own venue can print a price through your liquidation level without triggering it, if the cross-venue index hasn't moved there.
Because maintenance margin is tiered, large positions are typically liquidated in stages: the engine reduces your position to bring it down into a lower-requirement tier rather than dumping the whole thing at once. A $10M BTC position that breaches doesn't necessarily get fully closed — the system may cut it enough to restore margin at the next tier down. This is why the tiered tables aren't academic: they define both your liquidation price and how much gets closed when you hit it. Small positions (inside tier 1) are closed in one step.
Being liquidated is not free — a liquidation (or "clearance") fee is charged on top of the loss, which is why your realized loss at liquidation is worse than the raw price move suggests. On Binance, for example, the BTC-USDT contract carries a published liquidation clearance fee (0.0125 BTC-equivalent parameters appear in its public contract spec); every venue charges some form of it. Practically: your effective liquidation happens slightly before the theoretical price, because the fee eats into the maintenance buffer.
If your position is closed at a price worse than bankruptcy (your margin fully gone), the shortfall is absorbed by the venue's insurance fund — a reserve built up from liquidations that closed better than bankruptcy. A healthy, growing insurance fund means winning traders reliably get paid without clawbacks. This is a real point of difference between venues: the larger and more transparent the fund (most majors publish its balance), the less often the last stage triggers.
If the insurance fund can't cover a shortfall in a fast, one-sided move, the venue falls back to auto-deleveraging: it force-closes opposing positions, starting with the most profitable and highest-leverage traders on the other side, to make the counterparty whole. If you are a profitable short in a crash, ADL can close part of your winning position at the bankruptcy price without your consent. Venues rank each account's ADL exposure (often shown as an indicator in the position panel). ADL is rare on liquid majors and more common on thin alt contracts in violent moves.
None of this is exotic — it's the standard perpetual-swap liquidation model used across the six venues we track. The differences are in the parameters, and the parameters are knowable. Start with the maintenance-margin tiers and the live funding monitor for the two costs that move your liquidation price while you hold.