# Top Crypto Derivatives Exchanges, Ordered by the Depth Behind an Exit

Updated: 24 September 2026 | Book readings 22 September 2026 | The Margin Brief desk

## Summary

Ordered by the one figure produced here, the top crypto derivatives exchanges read on
22 September 2026 ran from edgeX, whose bitcoin book held
$29.3 million within 10 basis points of the mid, down to Paradex at $55.00.
The depth column and the score column disagree, and this page follows the depth.

Weighting: Exit liquidity 30 · Collateral and settlement 22 · Cost per round trip 20 · Loss record and audits 14 · Contracts and asset classes 9 · Getting a position on 5, fixed 22 September 2026. Method: https://insurancerefocused.com/how-we-rate

## What the figures decide

1. edgeX led on the only number produced here: $29.3 million resting within 10 basis points, and 0.22 bps of impact on a $1 million order.
2. Lighter holds the highest total on this site at 12.5 of 15 and comes third on this page, because the order here is the depth reading.
3. Hyperliquid reported $8.7 billion of open interest on 22 September 2026 and held $7.4 million; Extended reported $84.0 million and held $7.8 million.
4. The thin end is not a rounding error: Paradex held $55.00, where an order of $1 million found no price at all, against $7.4 million at Hyperliquid one rung above it.

## Seven venues in order of the depth behind an exit

The order is the depth column in Exhibit B, not the score column: the highest total on this site sits
third here. Posted collateral is the whole of what one of these contracts can lose.

| # | Venue | Score of 15 | Maker / taker | Markets | Model |
|---|---|---|---|---|---|
| 1 | edgeX | 11.5 | 0.040% / 0.045% | 172 | Hybrid |
| 2 | EVEDEX | 9.7 | 0.015% / 0.045% | 52 | Hybrid |
| 3 | Lighter | 12.5 | 0% / 0% | 214 | Hybrid, ZK-proven |
| 4 | Aster | 11.3 | 0% / 0.040% | 578 | Own layer 1, operator-run |
| 5 | Extended | 11.5 | 0% / 0.025% | 325 | Hybrid |
| 6 | Hyperliquid | 10.2 | 0.015% / 0.045% | 324 | On-chain order book |
| 7 | Paradex | 9.0 | 0% / 0% | 63 | Hybrid, Starknet appchain |

Venue facts as of 18 September 2026; entry-tier fees on the BTC perpetual. Access differs by country.

## Order-book readings

BTC perpetual books read over 122 polls between 21 September 2026 and 22 September 2026; medians of all successful snapshots, no orders placed.

| Venue | Spread | Depth, 10 bps | Impact, $100,000 | Impact, $1 million | Snapshots |
|---|---|---|---|---|---|
| edgeX | 0.04 bps | $29.3 million | 0.02 bps | 0.22 bps | 121 |
| EVEDEX | 4.12 bps | $18.4 million | 2.06 bps | 2.75 bps | 121 |
| Lighter | 0.01 bps | $16.9 million | 0.65 bps | 1.48 bps | 50 |
| Aster | 0.01 bps | $14.8 million | 0.44 bps | 1.11 bps | 122 |
| Extended | 0.12 bps | $7.8 million | 0.06 bps | 0.70 bps | 50 |
| Hyperliquid | 0.12 bps | $7.4 million | 0.06 bps | 0.72 bps | 122 |
| Paradex | 18.39 bps | $55.00 | 162.91 bps | not filled | 120 |

## Assessments

### 1. [edgeX](https://pro.edgex.exchange) — size that has to leave in a hurry

Score 11.5/15.

edgeX matched the largest resting size of the seven: a median $29.3 million within 10 basis
points, and 0.22 bps of impact on a $1 million order. Settlement and a withdrawal
that does not need the operator run on its own chain ([edgeX](https://pro.edgex.exchange/en-US/home), checked 18 September 2026).

- Works: The largest resting size of the seven; An exit route that does not depend on the operator.
- Falls short: 0.040% maker, the dearest maker rate among these seven; An EDGE flash crash on 2 June 2026 triggered liquidations, met with goodwill payments.
- Not for: traders who work orders on the maker side.

### 2. EVEDEX — crypto, metals, oil and shares on one balance

Score 9.7/15.

EVEDEX matches orders off-chain and settles on Arbitrum. Its bitcoin book held
$18.4 million within 10 basis points, second here and ahead of venues reporting far larger
figures. It lists 52 perpetual contracts on one cross-margined
balance posted in USDT, at 0.015% maker / 0.045% taker before cashback (EVEDEX documentation, our reading of 18 September 2026).

- Works: Second on the depth reading; Crypto, metals, oil, currencies and shares on one balance.
- Falls short: A $1 million order moved the mid 2.75 bps, more than at four venues holding less depth; 52 contracts against 578 at Aster, with no spot market, no dated futures and no bug bounty listed on CertiK Skynet.
- Not for: anyone needing a spot balance; not intended for UK retail clients.

### 3. [Lighter](https://lighter.xyz) — a round trip that costs nothing

Score 12.5/15.

The zero-fee Standard account at Lighter is priced in latency instead, at 300 ms on a taker order
([Lighter docs](https://docs.lighter.xyz/trading/trading-fees.md), checked 18 September 2026). Its book held $16.9 million,
third here.

- Works: Nothing to pay on the default account; Withdrawals can be forced through a priority queue on Ethereum.
- Falls short: A $100,000 order moved the mid 0.65 bps, wider than edgeX or Aster managed on this page; A 4.5-hour outage on 10 October 2025, with a published post-mortem and compensation.
- Not for: traders who judge a venue by its fee schedule alone.

### 4. [Aster](https://www.asterdex.com) — the widest list of contracts

Score 11.3/15.

Aster lists 578 perpetual markets on its own chain and quoted 0.01 bps,
the tightest median spread recorded here, level with Lighter. Behind that quote stood
$14.8 million, and a $100,000 order moved the mid 0.44 bps.

- Works: More perpetual markets than any venue here, across six asset classes; The tightest median spread recorded, 0.01 bps, level with Lighter.
- Falls short: On 6 October 2025 DefiLlama pulled its perpetual volume figures, saying it could not verify them; Chain core code is closed-source, with no external validators in phase one.
- Not for: traders who need a volume record verifiable from outside.

### 5. [Extended](https://extended.exchange/) — the widest reach for the money

Score 11.5/15.

Extended runs one flat schedule for every account, with no volume tiers, and lists
325 markets across six asset classes on Starknet. Its bitcoin book held
$7.8 million, a quarter of the reading at edgeX.

- Works: One flat schedule for every account, with no tier to climb; Bug bounty of up to $500,000 and no loss of user funds in 24 months.
- Falls short: 278 of its 325 markets have no public order book; Reported open interest of $84.0 million, the smallest among the six deepest books here.
- Not for: traders who need a visible book on every market they trade.

### 6. [Hyperliquid](https://hyperliquid.xyz) — the largest reported book here

Score 10.2/15.

Hyperliquid runs its book on its own chain, where every order, trade and liquidation is recorded.
CoinMarketCap put its open interest at $8.7 billion on 22 September 2026. The book read held
$7.4 million, sixth here.

- Works: The largest reported open interest on this page; 0.72 bps on a $1 million order, the third smallest here.
- Falls short: Sixth of seven on measured depth, on the largest reported open interest of the seven; About $4.9 million of vault bad debt in November 2025, and $4 million lost in March 2025.
- Not for: traders who read a reported open-interest figure as available size.

### 7. [Paradex](https://www.paradex.trade) — nothing to pay, little to trade against

Score 9.0/15.

Paradex charges retail orders no maker or taker fee, against a 300 ms speed bump and retail rate
limits ([Paradex docs](https://docs.paradex.trade/trading/trading-fees.md), checked 18 September 2026). Behind that price stood a median
$55.00 at a spread of 18.39 bps, and an order of $1 million found no
price at all.

- Works: No fee on retail orders since 15 June 2026; Bug bounty of up to 500,000 USDC.
- Falls short: The thinnest book on this page, by a wide margin; 162.91 bps of impact on a $100,000 order.
- Not for: any position that cannot be closed in pieces.

## Sources quoted

> "There is no public order book." — Extended documentation, request-for-quote execution, 18 September 2026. https://docs.extended.exchange/extended-resources/trading/rfq-execution

> "Orders are only stored on the Paradex Cloud and are only known to the operators of Paradex Cloud." — Paradex documentation, 18 September 2026. https://docs.paradex.trade/trading/privacy.md

> "HyperCore includes fully onchain perpetual futures and spot order books." — Hyperliquid documentation, 18 September 2026. https://hyperliquid.gitbook.io/hyperliquid-docs/about-hyperliquid.md

## What the depth reading changes about the top crypto derivatives exchanges

Reported open interest describes what a venue says has happened. Resting depth describes what is
available now. The same seven venues, ordered by the first against the second:

| Venue | Open interest reported | Median depth measured |
|---|---|---|
| Hyperliquid | $8.7 billion | $7.4 million |
| Aster | $1.4 billion | $14.8 million |
| EVEDEX | $704.9 million | $18.4 million |
| edgeX | $635.5 million | $29.3 million |
| Lighter | $585.4 million | $16.9 million |
| Extended | $84.0 million | $7.8 million |
| Paradex | $23.7 million | $55.00 |

The middle column is supplied by each venue and published by CoinMarketCap on 22 September 2026; nobody
outside it measures the figure. Hyperliquid reports about a hundred times what Extended does and held
slightly less depth than it; edgeX sits fourth on the reported figure and first on the measured one.

## A tight spread is not depth

Aster and Lighter quoted the tightest median spreads recorded here, 0.01 bps, and a
$100,000 order still moved the Aster mid 0.44 bps and the Lighter mid
0.65 bps. Hyperliquid and Extended quoted 0.12 bps; EVEDEX
quoted 4.12 bps on the second deepest book here and moved 2.06 bps.
A spread is what one lot costs; depth is what the position costs.

## How the figures were taken

Nobody publishes what an exit costs, so depth, spread and impact are this desk's own: every public BTC
perpetual book read over 122 polls between 21 September 2026 and 22 September 2026, no account anywhere, no order of ours. Charges, margin
demands, ladder rungs, market lists and audit history are each venue's own record on 18 September 2026.
Six weights, settled 22 September 2026 and unchanged since, carry 7 venues
([how we rate](/how-we-rate)).

The readings cover BTC only, and a smaller market is thinner. They are medians of a 91-minute window on
one day, not a property of any venue. Entry-tier rates ignore cashback and volume tiers, and reported
open interest is each venue's own figure rather than a measurement.

## FAQ

### How is order-book depth measured?

By adding up the dollar value of every resting bid and offer within a set distance of the mid price.
The band here is 10 basis points on each bitcoin perpetual book, read every ten minutes on
22 September 2026 and reported as a median.

### Does a bigger exchange always have a deeper order book?

No. Hyperliquid reported $8.7 billion of open interest on 22 September 2026 and held a median
$7.4 million within 10 basis points; edgeX reported $635.5 million and held
$29.3 million. A reported figure counts contracts already open, while depth counts orders
resting now.

### What is slippage on a perpetual futures order?

The gap between the price on the screen and the price the order actually paid. On
22 September 2026 a $1 million market order moved the mid
0.22 bps at edgeX and 0.70 bps at Extended; the Paradex
book could not fill it at all.

### Why do exchanges report different open interest figures?

Because each venue reports its own. CoinMarketCap published the figures on this page on
22 September 2026 as the venues supplied them: $8.7 billion at Hyperliquid,
$704.9 million at EVEDEX, $23.7 million at Paradex. Nobody outside the venue measures them,
so this page orders on depth.

### Do zero-fee crypto exchanges cost more in the end?

They can. Paradex and Lighter both charge nothing at the entry tier. A $1 million order met
1.48 bps on the Lighter book and no fill at all on the Paradex book, while
edgeX, at 0.040% maker / 0.045% taker, moved 0.22 bps.

### Does a tight spread mean deep liquidity?

Not on these readings. Aster and Lighter quoted the tightest median spreads recorded,
0.01 bps, yet a $100,000 order still moved the Aster mid 0.44 bps.
EVEDEX quoted 4.12 bps on a deeper book and moved 2.06 bps.
Quote and size are separate facts, and only one of them is the exit.

### Why measure depth within 10 basis points instead of the whole book?

Because a margined position is closed at market, and the orders furthest from the mid are the ones
it reaches last and pays most for. Ten basis points is a band close enough that an order inside it
barely moves the price: on this page it held $29.3 million at the top of the table and
$55.00 at the foot.

### What is a basis point in crypto trading?

One hundredth of one percent. On $1 million of exposure a basis point is $100, so the
1.48 bps a market order met on the Lighter book was about
$148.00, and the 0.22 bps it met at
edgeX was about $22.00. Slippage and fees are quoted the
same way.

### How often does order-book depth change?

Constantly. These readings are medians of up to 122 snapshots taken ten
minutes apart between 21 September 2026 and
22 September 2026. Two of the seven books answered barely half of those
attempts, so their medians rest on a thinner run. A median smooths a window; it promises nothing
about the book tomorrow.

### Can a market order fail to fill on a derivatives exchange?

Yes. On the Paradex book read here, orders resting within 10 basis points came to a median
$55.00, and an order of $1 million found no price at all. The order walks the book
until filled, at whatever prices are left standing.

### What happens to a leveraged position when liquidity disappears?

It is closed at whatever the book offers. At twenty times leverage, $1 million of exposure sits on
$50,000 of collateral, and on the Paradex book that order found no price at all inside the reading.
A position that cannot leave in one piece leaves in several, at whatever is left standing.

## Editorial note

These are leveraged perpetual futures: the margin behind a position is all of what it can lose. Placement here is paid for; the order of the table is not. Corrections: editorial@insurancerefocused.com.

The Margin Brief desk, 24 September 2026

Placement in this file is paid for; the order of every table follows the published formula and nothing else.