Before GMX, DeFi yields were mostly emissions: protocols printed tokens to pay liquidity providers, and the yield disappeared when the printing stopped. GMX distributed actual trading revenue in ETH, which reframed what a sustainable return looks like and started the real yield narrative.
The pool-as-counterparty model
Liquidity providers deposit into a pool that takes the other side of every trade. Traders get oracle-priced fills with no slippage; providers earn fees, borrow costs and, on average, trader losses. The design works because most leveraged traders lose over time — and it means providers are directionally exposed and can lose heavily in a sustained trend where traders are correct.
v1, v2 and the exploit
v1 used a single pooled basket. v2 replaced it with isolated GM markets, each with its own backing assets, which contains risk per market. In July 2025 a re-entrancy vulnerability in v1 on Arbitrum allowed an attacker to manipulate pool accounting and extract roughly $40m; much was returned following a bounty offer, v2 was unaffected, and v1 was deprecated.
Where it stands
v2 operates across Arbitrum and Avalanche with real fee revenue and transparent on-chain state. Competition from Hyperliquid and other order-book venues has taken volume, and the borrow-fee model still makes long-held positions expensive.
Who should care
Traders wanting slippage-free entries, liquidity providers evaluating whether to be the house, and anyone assessing which DeFi yields are actually funded by revenue.