GMX v2 liquidation mechanics — what CEX traders get wrong

By lz · Apr 14, 2026 · 7 min read

If you opened your first GMX v2 position after years on Binance or Bybit, the liquidation surface looks deceptively familiar. Same long/short, same leverage slider, same maintenance margin idea. It is not the same. GMX v2 prices its perps against GM pools — isolated, asset-specific liquidity vaults — and that single architectural decision changes how borrow, funding, and price impact interact with your liquidation. The formula you memorized on a CEX will be off, sometimes by a lot, and almost always against you.

This guide walks the full mechanism. By the end you will know which inputs the calculator on this page handles correctly, which it deliberately does not, and where to go look in the GMX UI to fill the gaps before you click trade.

The core formula

Strip away the GMX-specific machinery and the inequality is the one you already know. For an isolated long, you get liquidated when:

collateral − fees < positionSize × MMR

Where MMR is the market’s maintenance margin rate — about 0.56% on the BTC and ETH GM markets, tighter on long-tail markets like SOL or ARB. Solve for price, plug in your numbers, and you have the same closed-form expression that the calculator on this page evaluates:

Estimated Liquidation Price

$45310.00

-9.38%from entry

Estimates only. Actual GMX v2 liquidation also includes accrued borrow fees, funding fees, and price impact at fill-time — none of which are modelled here. Not financial advice.

That number is correct at the moment you open the position. Every subsequent second, three things drift it: borrow fees, funding fees, and the price impact you will eventually pay to close. None of them are in the formula. They are why the rest of this guide exists.

Borrow fees — the silent killer on skewed pools

GMX v2 charges per-second borrow fees on the larger side of an imbalanced GM pool. If long open interest dwarfs short on the BTC GM market, longs pay borrow continuously to compensate the LPs taking the other side of their directional exposure. The rate is small in any single second — typically a fraction of a basis point — but it is paid out of your position margin, not your wallet, and it accrues 24 hours a day.

Practical example: a 10× long held for a week against a heavily skewed pool can easily lose 1–2% of margin to borrow alone. Your formula-derived liquidation price drifts toward you by a corresponding amount the entire time, even if spot does not move. This is the most common way fresh GMX v2 traders get surprised — they hold a thesis trade for a few days, the chart looks fine, and the position quietly liquidates because their effective collateral has been bleeding.

The fix is not to avoid GMX. The fix is to check the borrow rate on the GMX UI before you open, and either size down or pre-budget the bleed into your stop-loss.

Funding fees — newer, smaller, but asymmetric

v2 introduced funding to GMX in addition to borrow. Conceptually it is the standard CEX-style payment between longs and shorts based on open interest imbalance, but in practice the rates are smaller than what you see on Binance and the asymmetry can be sharp. On a heavily one-sided market both sides can sometimes pay (one to LPs as borrow, the other to the imbalanced side as funding), and which way it cuts depends on the specific OI balance and the market’s parameters.

For most positions held under a day this is rounding error. For positions held over a week it stacks with borrow and is worth pricing in.

Price impact — open and close both bite

Every GM market quotes a price impact that scales with how much your trade shifts the pool toward or away from balance. Open against the skew (helping balance the pool) and impact is positive — you get a bonus on entry. Open with the skew (worsening the imbalance) and you pay impact, eating into the position before any market move at all.

The same logic runs in reverse on close. A long that helped balance the pool when it opened will pay impact to close it later, because closing it shifts the pool back toward imbalance. This is bidirectional friction that the static formula cannot model: your effective entry price is not the mid you see on the chart, and your effective exit price is not the mid at the moment of liquidation either. On a $50,000 position at typical GM market depth, total round-trip impact can easily be 20–50 bps. On a thinly traded market it can be much more.

Per-market MMR and leverage caps vary

The calculator on this page ships with the BTC/ETH GM market parameters: MMR ~0.56%, max leverage 100×. These do not apply uniformly across GMX v2. SOL, AVAX, ARB, LINK, and the other long-tail GM markets each have their own parameter sheets, and most cap leverage substantially lower (50× is common, 30× exists) and use higher MMR (1% or more is typical).

If you trade anything other than BTC or ETH on GMX v2, you must look up the market’s actual parameters in the GMX docs or live UI before trusting any calculator output, including this one. The math is correct; the parameters are not the parameters of the market you are about to enter.

Isolated vs cross margin

GMX v2 supports both. Isolated is the safer default: each position has its own assigned collateral, the liquidation price depends only on that collateral, and a wipeout on one position cannot touch another. Cross lets you draw against your full account equity, which pushes the liquidation price much further away — at the cost of making one bad trade capable of taking the entire account.

The calculator’s “Cross” mode exposes an Account equity input so you can see the real impact. The general caveats around cross margin are identical to those on Hyperliquid, so if you want the longer treatment of when cross goes wrong, see the Hyperliquid liquidation mechanics guide — the architectural differences do not change the human failure modes.

A concrete example

You open a $25,000 notional long on BTC GM at a $50,000 entry, 10× leverage, isolated, fees included.

  • Initial margin: $2,500.
  • BTC GM MMR: 0.56% → maintenance requirement $25,000 × 0.0056 = $140.
  • Round-trip fees at 3 bps each side on $25,000 = $15.
  • Buffer before liquidation: $2,500 − $140 − $15 = $2,345.
  • Price move that eats the buffer: ($2,345 / $25,000) × $50,000 = $4,690.
  • Liquidation price: $50,000 − $4,690 = $45,310 (−9.38% from entry).

Now hold that position for a week against a 0.05%/day borrow rate. You bleed $25,000 × 0.0005 × 7 = $87.50 of margin. Your effective buffer is now $2,257.50, and your liquidation price has drifted to about $45,485 — a $175 worse liquidation level with the chart unmoved. Add 30 bps of close-side price impact ($75) on top of that and the picture darkens further.

The static number from the calculator is correct on day zero. The day-seven number is something the calculator cannot give you, and that is exactly the gap CEX traders forget exists.

What the calculator doesn’t model (recap)

  • Borrow fees over the hold. Initial liquidation only — the largest source of drift on GMX v2.
  • Funding fees. Asymmetric and smaller, but real on multi-day holds.
  • Price impact. Both sides of the round-trip.
  • Per-market parameter variance. Defaults are BTC/ETH; long-tail markets need lookup.

For every one of these, the GMX live UI shows you the current state before you commit. Use it.

Last verified against GMX’s live UI on 2026-04-14. If GMX ships parameter changes we haven’t updated, send me a DM on X (see about).