Binance Cross vs Isolated Margin: What Changes at Liquidation?

The position is the same; the wallet is not

A BTCUSDT perpetual is one contract with one set of rules. What changes between cross and isolated margin is the collateral the venue is allowed to reach for when the position moves against you. Binance’s support documentation describes cross margin as sharing one margin balance across positions, and isolated margin as giving each position its own dedicated margin. The liquidation price is then calculated from that balance.

That distinction sounds administrative until it is expressed in dollars. On 2026-10-19 at 01:33 UTC the BTCUSDT perpetual marked at 82,477.50 USDT, with an index price of 82,496.42 and a last trade of 82,472.40. The contract’s exchangeInfo entry reported liquidationFee: 0.012500, a tick size of 0.10 and a minimum notional of 50 USDT. None of those values depend on the margin mode. The account balance behind them does.

The last two rows are operational, not theoretical. Binance’s documentation states that all contracts default to cross margin, that switching applies only to the selected contract, and that the mode cannot be changed while the contract has open orders or positions. A trader who intends to use isolated margin has to decide that before the order goes in.

What the liquidation calculation actually reads

Binance’s liquidation page names the terms rather than presenting a number: wallet balance (WB), maintenance margin (TMM), and unrealised profit and loss (UPNL). For cross mode, WB is the cross wallet balance. For isolated mode, WB is the isolated wallet balance of that position, and the page gives the substitution as TMM = 0 and UPNL = 0 for the other positions — meaning they drop out of the calculation entirely.

The algebraic formula on that page is published as an image, so this article does not reproduce it. What matters for comparing the two modes is the structure:

• Maintenance margin is position notional multiplied by a maintenance margin rate for the tier the notional falls into, minus a maintenance amount for that tier. It rises as the position grows.
• Wallet balance is the collateral the venue can apply. In cross mode that is the shared balance; in isolated mode it is the dedicated margin.
• Unrealised P&L moves with the mark price and either adds to or subtracts from the cushion.

A liquidation happens when the cushion between adjusted equity and maintenance margin reaches zero. In cross mode, every position’s unrealised loss is subtracting from the same WB. In isolated mode, only that position’s loss is.

A worked comparison on today’s data

The clearest way to see the difference is to hold everything constant except the mode. Assume a 0.5 BTC long entered at 82,000 USDT, marked at 82,477.50, with a maintenance margin rate of 0.4% and no other positions. The 0.4% is an assumption for the example, not Binance’s published tier — the venue’s bracket table is account-specific and not available from the public API. Fees other than the liquidation fee are ignored.

Position notional at the mark is 0.5 × 82,477.50 = 41,238.75 USDT. Maintenance margin at the assumed rate is 164.96 USDT. Unrealised profit is 0.5 × (82,477.50 − 82,000) = 238.75 USDT.

Isolated, with 3,000 USDT assigned to the position. The position’s equity is 3,000 + 238.75 = 3,238.75. The cushion above maintenance margin is 3,238.75 − 164.96 = 3,073.79. A fall of that much across 0.5 BTC is 6,147.58 per BTC, putting the simplified liquidation price near 76,329.92.

Cross, with 10,000 USDT in the wallet and no other positions. Equity is 10,000 + 238.75 = 10,238.75; the cushion is 10,073.79; the same arithmetic gives a fall of 20,147.58 per BTC and a liquidation price near 62,329.92.

The position did not change. The entry, size, mark and assumed maintenance rate are identical. The liquidation price moved by roughly 14,000 USDT because the wallet behind it moved.

Now add a second position. Suppose an ETH long in the same cross wallet is carrying 2,000 USDT of unrealised loss. Cross wallet balance available to the BTC position falls by 2,000, so the BTC liquidation price rises by 2,000 ÷ 0.5 = 4,000 USDT, to roughly 66,330. If the ETH position is isolated, its loss stays in its own wallet and the BTC position’s liquidation price does not move. That is the whole trade-off: cross margin gives each position a deeper cushion, at the price of making every position share the same fate.

The fee that estimates leave out

The contract’s liquidationFee field is 0.0125 on BTCUSDT, or 1.25% of notional. At 0.5 BTC and the mark above, that is 515.48 USDT. It is not a rounding error next to the 164.96 USDT maintenance margin in the example. Any liquidation price calculated from maintenance margin alone will be optimistic — the position reaches the venue’s threshold before the simplified number suggests, and the fee then lands on what is left.

This is also why two calculators can disagree by thousands of dollars. One uses a flat maintenance rate and ignores the fee; another uses a tier table that may not match the account’s actual bracket; a third assumes the last price rather than the mark. The exchange’s own liquidation estimate is the only one computed from the account as it stands.

What actually changes in practice

• Entry decisions. Margin mode is fixed per contract and cannot be changed while orders or positions are open. Choosing it is part of the trade, not a setting to revisit when the position is in profit.
• How much collateral is committed. Isolated margin makes the commitment explicit: the number you assign is the most that position can lose, plus its own unrealised profit. Cross margin does not cap anything; it pools.
• What a hedge protects. A hedge held in the same cross wallet offsets unrealised P&L inside one balance. A hedge in a separate isolated wallet offsets price exposure but not margin calls. The two are not equivalent, even when the net delta is zero.
• How a stop interacts with liquidation. A stop order is an instruction to trade; a liquidation is the venue closing the position when margin runs out. They are different mechanisms and can both be resting at once. The guide to OCO, OTO and trailing stops covers how protective exits are structured; the liquidation engine does not read your stop.
• How position size is set. Sizing from a risk budget assumes a stop distance you choose. Sizing for isolated margin also has to assume the margin you assign, because that, not the stop, determines when the venue steps in. The position-size calculator works the first calculation; the second is a separate decision about how much collateral to fence off.

Some of this is familiar from other venues. OKX’s account modes make a similar split between a shared account and isolated positions, covered in OKX account modes. The vocabulary is shared; the tier tables and fee fields are not, so numbers do not transfer between exchanges.

Risk note. All market values in this article were read from Binance’s public futures API on 2026-10-19 between 01:32 and 01:33 UTC, and the contract fields from exchangeInfo at the same time. The maintenance margin rate of 0.4% is an illustrative assumption, not Binance’s published schedule; the worked liquidation prices are simplified and ignore fees other than the one named, funding, and tier changes as the notional moves. A leveraged perpetual can lose the entire margin assigned to it and, in cross mode, more than the collateral you intended to risk on one trade. Nothing here is investment advice.

The two questions to ask before choosing

The first is mechanical: where does the liquidation price sit under each mode, given the collateral I am actually willing to commit? That is answerable before entry, with the venue’s own estimate, and it is the number that matters once the position is open.

The second is structural: if another position in the same wallet goes wrong, do I want it to be able to take this one with it? Cross margin answers yes. Isolated margin answers no, and charges for that answer by giving the position a shallower cushion.

Neither answer is universally right. What is not defensible is choosing a mode by habit and then estimating the liquidation price as if the other mode’s collateral were behind it. The position is the same contract either way; the wallet is the variable.