RISK Updated 2026-09-12 By Ivo Renner

Leverage tiers and liquidation on a perpetual: how it actually triggers

The number in the leverage selector is a ceiling for the smallest bracket of position size, and it falls away as your position grows — a real position rarely gets it. Almost everyone who is surprised by a liquidation was surprised by this first.

Leverage is a ladder, not a number

The board says CATUSDT offers 25×. That is true, and it is also the single most misleading fact on the page.

Binance does not apply one leverage limit to a contract. It applies a tier table: a series of position-size brackets, each with its own maximum leverage and its own maintenance margin rate. The headline number — the one in the announcement, the one on our contract list — is the figure for the smallest bracket only. As your position notional crosses each threshold, two things happen simultaneously and automatically:

• Maximum leverage steps down. The bracket that allows 25× might cover the first slice of notional; the next bracket might allow 10×, the one after that 5×, and so on down to 1× for the largest positions.
• Maintenance margin rate steps up. The percentage of notional you must keep in the account to avoid liquidation increases with every tier. Larger positions are held to a stricter standard.

The consequence people do not expect: adding to a winning position can move you into a worse tier and bring your liquidation price closer, even though you just added margin. You did not do anything wrong at the moment of the top-up; the ladder simply re-graded the whole position under new rules.

Where to read your contract’s actual ladder

This site refuses to print leverage tiers or maintenance-margin rates, because they are per-contract, they change after listing, and a stale number here would corrupt every calculation you built on it. That refusal is only defensible if we tell you where the real numbers live. Here they are, on a public page, no account needed.

1. Open Binance Futures and find Market Info. The public data section sits in the left sidebar of the futures interface. You do not have to be logged in to reach it — the capture below was taken logged out, with the Log In and Sign Up buttons still showing.
2. Choose Contract Specifications. This is the sub-page carrying per-symbol values. Trading Parameters nearby carries the order-mechanics limits (minimum quantity, price precision, order caps) which matter for a different reason: they decide whether the quantity your position sizing produces is actually placeable.
3. Search your symbol. Type the contract, not the ticker: TSLAUSDT, not TSLA.
4. Read the bracket your notional falls into. The ladder is a table, not a single figure. Find the row whose notional range contains your intended position, and take the maximum leverage and maintenance-margin rate from that row — not from the headline number in the leverage selector.
5. Cross-check against the order form. Before confirming, the order ticket shows an estimated liquidation price. If it disagrees with what you calculated, trust the exchange and work out why afterwards.

The columns you are looking for, and what each one does to you:

What actually happens as a position moves against you

A liquidation is not an event. It is the last step of a sequence, and each step is observable before the one after it.

1. The mark price moves against you. Your position accrues an unrealised loss. Nothing has been realised, nothing has been charged, but the number Binance uses to value your position has changed.
2. Your margin ratio rises. Margin ratio is, broadly, maintenance margin divided by margin balance. As unrealised loss eats the balance, the ratio climbs. This number is displayed in the interface. It is the single most useful thing on the screen and most people never look at it.
3. Funding and fees compound the drift. At each funding settlement shown for that contract, if you are on the paying side, funding is deducted and the ratio worsens a little more. The interval is contract-specific. See funding on equity perps for what that costs.
4. Margin balance reaches the maintenance margin requirement. The ratio hits 100%. This is the liquidation trigger — not zero equity, not "you lost your money", but the point at which your remaining balance equals the minimum the exchange requires to keep the position open.
5. The liquidation engine takes over. Binance closes the position. Depending on size and tier it may reduce the position in stages rather than all at once. What you get back, if anything, is whatever remains after the close is completed.

Step 4 is where the intuition breaks. You are not liquidated when your loss equals your margin. You are liquidated before that, when your margin has fallen to the maintenance requirement, which is a positive number. The maintenance margin is the exchange's buffer against the position going further underwater during the close — and it comes out of your side of the ledger.

Liquidation reads the mark price, not the last trade

This is the second thing people get wrong, and on TradFi perpetuals it matters more than it does on crypto.

Your liquidation is triggered off the mark price, not the last traded price on the chart. During regular hours Binance computes the mark as Median(Price 1, Price 2, Contract Price), recalculated every second. A median of three inputs is deliberately resistant to any single one of them, which is exactly what you want when a liquidation engine is reading it: a one-second wick on thin liquidity should not close your position.

The corollary is uncomfortable. The candle on your screen is not the number that decides your fate. A contract can print a trade well below your liquidation level and leave you untouched, because the mark did not follow. It can also, less often, move against you while the last trade sits still.

On a TradFi perp there is a further wrinkle: when the underlying cash market is shut, fresh vendor quotes may be unavailable, but the current equity-perp index does not freeze. Binance uses Orderbook EWMA mode, deriving the index from the contract book's impact mid price and smoothing it. Because the index is in Orderbook EWMA mode, the regular mark-price median still applies; EWMA smoothing of the futures last price is reserved for periods in which an underlying type's index is actually in Fixed mode. The full construction is set out in the price index and mark price guide, and it is worth understanding before you hold a leveraged position through a Friday close.

The exit you were relying on is an order, and orders have quotas

Everything above assumes you intend to be out before the engine gets there, and the usual plan for that is a stop. But a stop is an order, it lives in the same order system as everything else you place, and that system publishes limits on how many you may leave resting and where you may put them. They are printed on the public parameters page this guide has already sent you to twice — not in the table, but in the two lines of small print above it.

Three things are stated there, and the page gives them without comment:

• Conditional orders are capped at 200 open orders per user, counted across all symbols rather than per contract.
• Stop Limit, Stop Market and Trailing Stop orders are capped at 10 open orders per symbol, with a total of 10,000 open orders across all symbols.
• The table carries a column headed Limit Order Price Cap / Floor Ratio, and on every row visible it reads 5% / 5% — a bound on how far above and below the reference price a limit order may be priced at all.

The page uses two categories, “conditional orders” and “regular Stop Limit, Stop Market, and Trailing Stop orders”, and does not say there where the line between them falls. Until you know which bucket your order type lands in, assume the tighter number applies.

The first two are the ones nobody plans around, because protection consumes quota. A trader who exits in thirds (three stops staggered below the entry, two take-profit orders above it) has spent half a symbol's allowance on one position. Add a second position in the same contract, scaled or hedged, and the allowance is gone. Ten is generous for a clean trade and tight for a laddered one, and the way you learn which you were running is that the eleventh order is refused. The account-wide figure has the same shape one level up: it is met not by your worst position but by the sum of all of them, without warning, at the moment you add the leg that crossed it.

The price cap and floor is a constraint of a different kind: not how many orders, but where a limit order may sit relative to a reference. The screenshot rows above showed a 5% / 5% ratio for four crypto perpetuals in August 2026; they do not establish the current ratio, reference price or treatment of a TradFi symbol. Read the selected symbol’s Contract Specifications or Trading Parameters and any validation shown before submission. If the interface does not disclose the boundary in a read-only state, treat it as unknown. Do not submit a test order to discover it.

Two consequences feed back into the arithmetic on this page. A stop that fires into a market order is a taker order, and so is a forced close, which makes the exit the expensive leg — priced as though you were in a hurry, because you are; the cost guide prices it. And a stop is not a race the exchange owes you a win in: if your stop triggers on the last price while your liquidation reads the mark, the two are watching different numbers and can fire in an order you did not intend. The mark price guide sets out how far apart those numbers can sit.

Tesla at 5×, Caterpillar at 25×: reading the roster

Leverage is one of several specification fields Binance varies across the TradFi roster. Funding intervals and caps can also differ: nine TradFi contracts moved to four-hour funding with a ±1% cap on 10 September 2026. Read the live specification for the symbol alongside its leverage tier; the leverage cap alone is not a complete risk grade.

The extremes are instructive. TSLAUSDT, listed on 28 January 2026 as the first single stock on the platform, carries 5× — the lowest cap on the roster. CATUSDT, listed 2 July, carries 25×. Caterpillar is not five times safer than Tesla in any risk model anyone would recognise.

Two things are being expressed at once, and it pays to separate them.

Instrument risk. Within a single batch, the differences do look like genuine grading. In the 9 July 2026 batch, most contracts launched at 25× — including two leveraged ETFs, INTWUSDT (GraniteShares 2× Long INTC) and SNXXUSDT (Tradr 2× Long SNDK). But two microcaps in the same batch, BNCUSDT and FWDIUSDT, were held to 10×. Binance looked at those two names, in a batch where the default was 25×, and cut the ceiling by more than half. That is a deliberate statement about liquidity and volatility in the underlying.

Platform maturity. The other pattern is chronological. January's first single stock got 5×. The March and April batches — Meta, NVIDIA, Alphabet, QQQ, SPY, Apple, TSM — got 10×. June's Korean listings and the June US batches got 20×. The five July batches defaulted to 25×. The roster's leverage caps rose as Binance grew comfortable with the product, not because the underlying equities became safer — and, as the next section shows, the rise did not continue.

The practical reading: a low cap is a warning, but a high cap is not a reassurance. When Binance holds a contract to 10× in a 25× batch, it is telling you something specific about that underlying. When it grants 25×, it is mostly telling you the product line is eight months older than it was in January.

A leverage cap has a date on it

Those two readings pull against each other on any individual contract, and there is a way to tell which one you are looking at. It is the launch date, and the 10× band shows why.

Eleven contracts sit at 10×, and they are not one group. Seven are METAUSDT, NVDAUSDT, GOOGLUSDT, QQQUSDT, SPYUSDT, AAPLUSDT and TSMUSDT, listed across two batches in March and April and about as liquid as listed equity gets. Two are BNCUSDT and FWDIUSDT, microcaps held to 10× on 9 July in a batch whose default was 25×. The remaining pair, FLNCUSDT and RKLBUSDT, came on 18 May and are the subject of the paragraph after this one. The cap is identical and it means the opposite thing. On Apple it is a fossil of a cautious spring. On the microcaps it is a live judgement, made in the month Binance was granting 25× to nearly everything else it listed.

So the field is only readable in pairs: cap and launch date, never cap on its own. A cap below its own batch's default is a statement about the underlying; a cap equal to it is a statement about the calendar. Our contract list prints the listing date in the column beside the leverage figure, so the comparison takes a glance rather than a search through announcements.

Applied across the dataset, that test turns up a second graded batch that is easy to miss. On 18 May three contracts launched together: DRAMUSDT, a memory-sector ETF, at 20×, and FLNCUSDT and RKLBUSDT, two single names, at 10×. The diversified basket got double the ceiling of the single stocks announced beside it, in the same notice, on the same day — the July microcap cut running in the other direction. The pair is the clearest evidence on the roster that within-batch grading is about the instrument rather than the date.

The calendar reading, meanwhile, is not the straight line up that this page used to imply. Leverage climbed from 5× in January to 10× in the spring to 20× through June, and all five July batches defaulted to 25×. Then it came back down. Every contract listed in August 2026 launched at 20× — the two blue chips on 6 August, the six infrastructure and power names on 17 August, and the five memory-cycle listings on 25 August, three of which are leveraged ETFs. On the evidence in front of us, 25× was a July phenomenon rather than a new baseline, and a reader who arrived in August and assumed the ladder only goes up would have got the direction wrong.

Counted across the whole roster the picture is plainer than the headline suggests. 20× is the modal cap, carried by 34 of the 72 contracts. 25× covers 24, 10× covers 11, and 5× covers exactly one. The two January commodity contracts carry no leverage figure in our dataset, because the announcement we sourced them from does not state one and we will not fill a gap with an assumption. So the number everybody quotes describes a third of the roster, and the median contract on this platform is a 20× contract — every one of those counts being a launch-day maximum, not a live reading of a tier table that Binance revises without announcing.

The adverse move each setting can absorb

Here is the arithmetic that should inform your leverage choice, on a 5,000 USDT notional position. The initial margin is notional divided by leverage. The adverse move that consumes all of it is simply the reciprocal of leverage, expressed as a percentage of notional.

The fourth column is the one to sit with. The formula is (1 / leverage) − m, where m is the maintenance margin rate on your bracket, before fees and funding. Liquidation always happens before the full move in column three — maintenance margin bites first, and the more leverage you use the larger a share of your buffer it takes.

Put concrete equity moves against those thresholds and the picture sharpens. A 4% single-day move in a large-cap stock is a normal Tuesday. A 10% move is an earnings reaction. A 20% move is a bad earnings reaction, and on a single name it is not rare. At 25×, an ordinary session in the wrong direction is enough. At 5×, it takes an event.

And remember that on a perpetual you are exposed to those moves continuously. The stock gaps on Monday's open, but your position has been marked against a smoothed index all weekend and will be re-marked into the gap whether you are awake or not. The gap exposure calculator exists for exactly this.

What a calculator estimates, and what actually happens

Our liquidation price estimator takes entry price, leverage, direction and margin, and returns a price. It is useful for sizing and for sanity-checking a plan before you place an order. It is an approximation, and the exchange's own figure governs. We say that plainly because the gap between the two is not academic.

Four things move the real number away from any static estimate:

• Funding accrual. Every settlement you pay through reduces your margin balance and pulls the liquidation price closer. A position that was comfortable on Monday is measurably less so by Thursday, with the price unchanged. The funding cost calculator quantifies the drift.
• Trading fees. Opening and closing both cost something, and the closing cost is charged against a position that is already in trouble. We do not publish fee percentages because the rate is tier-dependent; read your own row on Binance's public fee page and see how to price a trade.
• Tier changes. If you add to the position and cross a bracket threshold, the maintenance margin rate changes and so does the liquidation price. A calculator that assumed one tier is now describing a position that sits in another.
• The insurance fund and the close itself. The liquidation price is where the process starts, not the price you get. Binance's liquidation engine closes the position into the market; in fast conditions it may fill worse than the trigger. The insurance fund absorbs shortfalls beyond your margin, which is what stops a negative balance in most cases — but it is not a promise that you keep anything above the maintenance margin.