Paradex Review

How Paradex scores in perpetual DEXs

This entry rates Paradex, the perpetual futures exchange on its own Starknet-based appchain, in our perpetual DEX ranking. The basis is desk research of the Paradex documentation, its terms and its public GitHub organization, checked on November 1, 2026. No account was opened and no trades were placed. The category uses five criteria, each marked 0 to 10 and weighted equally, as set out in our ratings methodology.

The overall score is the plain average: (9 + 9 + 2 + 8 + 8) ÷ 5 = 36 ÷ 5 = 7.2. On the current peer scores that puts Paradex fifth of nine, 0.2 points behind Orderly Network and 0.2 ahead of Hyperliquid.

This check re-scored four marks from current sources. Fee transparency rose from 6 to 9 after the fees page set out the full retail and pro schedule. Funding and liquidation clarity rose from 2 to 9, audit disclosure from 2 to 8 and risk documentation from 2 to 8 after the funding, liquidation, audit and socialized-loss pages, unreachable at the previous scoring, were published. The overall score moves from 2.8 to 7.2. Licence and code openness is unchanged at 2.

Strongest criteria: fees and funding rules

Fee transparency (9)

The fees page splits orders into two lanes. Orders from the web app, or from the API with an interactive token, count as retail and pay 0% maker and 0% taker on perps, spot and options; Paradex's launch post dates this to September 16, 2025. Standard API orders count as pro. They pay a flat 0.003% maker fee and a taker fee that falls from 0.045% under $1M of 14-day volume to 0.035% above $25M. Three taker discounts add together: up to 30% for staking DIME, 40% for paying fees in DIME and 30% for FastFills. Pro taker fees cannot fall below 0.0175%. A pro taker at the entry tier trading $100,000 of notional pays $45 before discounts. The mark stops short of 10 because the maker rate's effective date is given without a year, and the DIME discount is described as an initial-rollout rate.

Funding & liquidation clarity (9)

Funding accrues continuously through an index and is quoted per eight hours. The premium is a weighted median of impact premiums from Paradex, Binance, Bybit, OKX, Hyperliquid and Lighter, with Paradex weighted 3.5 against 1.2 for the others. It is pulled toward a 0.01% baseline, clamped at ±0.05% and capped at 2% for BTC, ETH and SOL, 0.5% for TradFi markets and 5% for other perps. The liquidations page states the trigger (margin ratio above 100%), the 20% reduction steps that bring it below 90%, and a penalty of 70% of the maintenance margin on the share liquidated, paid to the insurance fund. The rules are stated as numbers a trader can work with; the mark is 9 rather than 10 because the per-market maintenance margins sit on a separate margin page rather than with the liquidation rules.

Audit disclosure (8) and risk documentation (8)

The audits page links reports from four named firms with dates. The core contracts (Paraclear, vaults, factory, registry and oracle) have one audit, by Cairo Security Clan in May 2025; Zellic and Trail of Bits reviewed the Ethereum bridge, and HackerOne tested the API and web app. A single core audit is why the mark is 8, below venues with several core reports. On risk, Paradex does not use auto-deleveraging. If the insurance fund cannot cover bankrupt accounts, a socialized-loss factor applies to withdrawals, so only users who withdraw during the shortfall take a haircut. A time-based "last resort" mechanism covers positions the fund cannot unwind, and the security pages describe operator controls that can block deposits and limit withdrawals. The fund's size is not published, which keeps this mark at 8.

Weakest criterion: closed core code

Licence & code openness (2): a current deficiency

The rubric reads the licence on the core protocol repository, not the SDK. Paradex's GitHub organization lists 24 public repositories. The original ones are SDKs, a CLI, an MCP server and documentation, mostly MIT-licensed, and its bridge contracts are a fork of StarkWare's StarkGate under Apache-2.0. None contains Paraclear, the contract that holds balances and settles positions, and order matching runs off-chain in what Paradex calls Paradex Cloud. This is a measured absence, not an unchecked one, and it is the reason the overall score is not higher.

What the score does not cover

• Liquidity and execution. Depth, spreads and fill quality, including the 300-millisecond speed bump on retail orders, were not measured.
• Privacy design. Encrypted on-chain state and the privacy council that holds decryption keys sit outside the five criteria.
• Access. US and Canadian persons are excluded by the restricted persons page. Availability is not scored.
• Bridges. Deposits from other chains go through third-party bridges with their own risks; see our guide to how bridges work.
• Operating history. Incidents are not a criterion. The terms list transaction rollback after server errors among the service's risks; no incident is described here without an official Paradex record we could open.

Paradex vs Hyperliquid, dYdX and GMX

Hyperliquid scores 7.0, 0.2 points below, with almost the same shape. Both mark 9 for fees and 8 for risk documentation, and both are held down by closed core code, with 2 each for licence. Hyperliquid's 10 for funding and liquidation, against Paradex's 9, is outweighed by Paradex's 8 for audits against its 6. dYdX (8.6) and GMX (8.4) sit above. dYdX's lead comes from its AGPL-licensed core, a 10 against Paradex's 2, while Paradex leads it on audits, 8 against 7. GMX marks 8 for licence and 9 for audits. Ostium, listed first in the category, also scores 8.6.

Who this rating suits

The score suits traders who want fee, funding, liquidation and loss rules published as figures before opening a position, and who accept a closed clearing contract. Readers who weigh a public, readable core codebase most heavily should look at the licence mark first. Leveraged perpetuals can lose more than a margin deposit; the rating describes documentation and is not advice to trade.