Skip to main content
CoinPulse
14 ranked

Best perp DEXs: 14 perpetual futures exchanges ranked on open data

Short answer: On data collected October 19, 2026, GMX ranks first of 14 decentralized perpetual futures exchanges at 7.04/10, ahead of Aevo (6.68) and Hyperliquid (6.6). Every entry has 100% of its score measured. The score rewards value locked, published audits and time in operation, which is why Hyperliquid, with by far the most value locked ($6.78B), ranks third rather than first: DefiLlama lists no published audit for it. Two things matter before you rely on this list: the perp DEX score does not include hack history, and it does not measure fees, funding rates, leverage or liquidity. A high score is not a statement that a venue is safe.

Data collected October 19, 2026 · Page checked October 23, 2026 · No affiliate links or paid placements

The full ranking

1GMX7.04100%
2Aevo6.68100%
3Hyperliquid6.6100%
4Gains Network6.24100%
5tieJupiter Perps6.16100%
6tieMUX Protocol6.16100%
7dYdX5.84100%
8Ostium5.76100%
9Drift5.56100%
10ApeX Protocol5.54100%
11Extended5.46100%
12edgeX5.26100%
13Paradex5.2100%
14Synthetix4.86100%

Click a column to sort. Names jump to the service’s card below; how each column is scored.

Best by documented use case

  • Highest overall evidence score: GMX is the only venue combining above-median value locked ($213.0M against a $30.5M median) with a published audit link on DefiLlama.
  • Most capital on the venue: Hyperliquid at $6.78B, then Jupiter Perps ($745.0M). Neither has an audit listed on DefiLlama, which leaves each 1.04 points behind a venue with one listed audit link. Large value locked shows the market's trust; it is not a code review.
  • Audited and long-running: Gains Network (4.9 years, 1 audit link, 5 chains) and Drift (4.9 years, 1 audit link). Read Drift's 2026 incident below before treating its audit count as reassurance.
  • Longest operating history: dYdX, 8.0 years since its 2018 launch. The value locked scored is for its current version (dYdX v4), so the age and the capital refer to different generations of the product.
  • If you need a status page: no perp DEX publishes a machine-readable incident feed the probe could read. Aevo, Jupiter Perps, dYdX, Paradex and Synthetix have a page you can open; the rest have none found.

What separates the scores

Scale of value trusted (30%)

Scored on a log scale. Hyperliquid (9), GMX and Extended (7) and Jupiter Perps (8) gain most here. Drift and Synthetix score 3. Synthetix is now measured as the whole protocol, summing all its DefiLlama listings: $883,499. Its V3 listing alone showed $0, a tracking gap.

Security audits (26%)

The sharpest divide. Six venues have one published audit link on DefiLlama (6/10): GMX, Aevo, Gains Network, MUX Protocol, Ostium and Drift. Eight have none (2/10), including Hyperliquid, Jupiter Perps and dYdX. No perp DEX scores higher than 6 on this snapshot. The criterion counts the audit links DefiLlama lists, not the number of audits a protocol has had, and it does not grade their scope or findings.

Track record (18%), multi-chain reach (10%), latency (8%), status transparency (8%)

These rarely change the order on their own. The youngest venues (edgeX and Paradex at 1.1 years, Extended at 1.6) score 6 on track record. Every venue scores 9 or 10 on latency. Status transparency is 2 or 5 for all fourteen.

How the perp DEX score is calculated

Criterion Weight Source
Scale of value trusted 30% DefiLlama TVL, log scale
Security audits 26% DefiLlama audit links (count of published links)
Track record 18% DefiLlama listing date (a floor) or documented launch
Multi-chain reach 10% DefiLlama chain count
Latency 8% CoinPulse probe, live interface, single location
Status transparency 8% Public status page on Statuspage, incident.io or Instatus

Not scored for perp DEXs: security record (hack history), incident repair time and peg stability.

Worked example: GMX = 7 × 0.30 + 6 × 0.26 + 9 × 0.18 + 8 × 0.10 + 10 × 0.08 + 2 × 0.08 = 2.10 + 1.56 + 1.62 + 0.80 + 0.80 + 0.16 = 7.04.

A criterion that cannot be measured is not counted as zero: its weight is spread across the others and "% measured" falls. A ranking place needs at least 25% of the weight measured. All 14 venues are at 100% on this snapshot. The methodology lists the disqualifiers that remove a service from every ranking; no perp DEX is excluded on this snapshot. Drift and Ostium were checked against the “funds never returned” disqualifier and kept, because restitution plans are in progress; the check is published in the disqualifier checks section of this page.

Incidents the score does not count

Because security record carries 0% weight in this category, hacks do not lower a perp DEX's score. DefiLlama's public hack database (checked October 23, 2026) lists these entries against ranked venues or earlier versions of them:

Venue Date Type Amount listed Note
Drift April 1, 2026 Proxy upgrade hijack $295M No returned amount recorded
Ostium July 15, 2026 Private key compromise $23.75M Also noted on the Ostium rating
GMX (V1) July 9, 2025 Reentrancy $42M $40M listed as returned
GMX (V1) September 18, 2022 Market manipulation $565K —
dYdX (V3) November 18, 2023 Risk parameter abuse $9M V3, not the rated v4
dYdX (V3) July 23, 2024 DNS hijack $31K —
Hyperliquid June 15, 2023 Risk parameter abuse $37K —
Synthetix (v1+v2) June 25, 2019 Oracle misconfiguration Not stated Older version

Source: DefiLlama hacks. Open each entry for context; a listing is a record of a reported incident, not a finding about who was at fault. If incident history matters to you, and for leveraged trading it should, weigh this table alongside the score.

What else the scores leave out

  • Trading costs: maker/taker fees, funding rates, spreads, slippage and liquidation penalties. The site does not publish fee data it has not verified through real trades.
  • Liquidity and leverage: order-book depth, open interest, maximum leverage and the markets listed.
  • How the venue is run: oracle design, liquidation engine, admin keys, upgrade controls and how much of the stack is on-chain. Several of the incidents above were key or upgrade-control failures that audit counts do not capture.
  • Access: KYC, restricted jurisdictions and frontend geoblocking. Check each venue's terms before connecting a wallet; for the US regulatory picture see our CFTC perps coverage.
  • UX and support.

Looking for centralized futures exchanges instead? The crypto exchanges ranking covers centralized venues on a different weighting; it does not score derivatives products specifically. For spot trading on-chain, see the DEX ranking.

How to choose a perp DEX

  1. Decide what you will trade (crypto majors, long-tail tokens, FX or commodities) and confirm the venue lists it with enough depth.
  2. Check the audit links and the incident table above. An audit is a starting point, not insurance.
  3. Look at value locked as a liquidity proxy, then check actual open interest for your market on the venue.
  4. Read the fee and funding pages, including liquidation fees, before opening a leveraged position.
  5. Confirm you are allowed to use it from where you live.
  6. Size positions you can afford to lose. Perpetual futures use leverage, and liquidations can close a position fully. This page is not investment advice.
  7. Keep custody basics in place: see self-custody crypto and how to avoid crypto scams.

New to the concept? Start with what a DEX is and CEX vs DEX.

Each service, and why it placed there

1

GMX

Not published
7.04100% measured

Latency 10/10. Median 146 ms to first byte from the live interface, our own probe, slowest sample discarded, single location.

Status transparency 2/10. No public status page found.

Not for: Anyone who needs to know when it breaks: no status feed, so you find out it is down by discovering it yourself.

Every criterion, with evidence
SCALEScale of value trusted
30% weight7/10

$213.0M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight6/10

1 published audit with links (DefiLlama) — read the findings yourself. Audits reduce risk, they do not remove it.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight9/10

Live 3.2 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight8/10

Deployed on 4 chains (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight10/10

Median 146 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight2/10

No public status page found. Nothing about its reliability can be verified from outside — including by us.

measured 2026-10-19 14:39 UTC
GMX full scorecard
2

Aevo

Status page, no readable feed
6.68100% measured

Latency 10/10. Median 106 ms to first byte from the live interface, our own probe, slowest sample discarded, single location.

Scale of value trusted 5/10. $14.7M in value locked (DefiLlama).

Every criterion, with evidence
SCALEScale of value trusted
30% weight5/10

$14.7M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight6/10

1 published audit with links (DefiLlama) — read the findings yourself. Audits reduce risk, they do not remove it.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight9/10

Live 3.5 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight8/10

Deployed on 4 chains (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight10/10

Median 106 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight5/10

A status page exists but publishes no feed we could parse — readable, not auditable. Part of that may be our blind spot (we read three platforms).

measured 2026-10-19 14:39 UTCevidence ↗
Aevo full scorecard
3

Hyperliquid

Not published
6.6100% measured

Latency 10/10. Median 56 ms to first byte from the live interface, our own probe, slowest sample discarded, single location.

Security audits 2/10. No published audits on DefiLlama.

Not for: Anyone putting size into it: it holds funds in smart contracts with no published audit. Fine to experiment with, not to trust with a position you cannot afford to lose.

Every criterion, with evidence
SCALEScale of value trusted
30% weight9/10

$6.78B in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight2/10

No published audits on DefiLlama. Not proof of insecurity, but an unaudited contract holding funds is a red flag.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight9/10

Live 3.5 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight8/10

Deployed on 2 chains (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight10/10

Median 56 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight2/10

No public status page found. Nothing about its reliability can be verified from outside — including by us.

measured 2026-10-19 14:39 UTC
Hyperliquid full scorecard
4

Gains Network

Not published
6.24100% measured

Latency 10/10. Median 78 ms to first byte from the live interface, our own probe, slowest sample discarded, single location.

Status transparency 2/10. No public status page found.

Not for: Anyone who needs to know when it breaks: no status feed, so you find out it is down by discovering it yourself.

Every criterion, with evidence
SCALEScale of value trusted
30% weight4/10

$8.7M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight6/10

1 published audit with links (DefiLlama) — read the findings yourself. Audits reduce risk, they do not remove it.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight9/10

Live 4.9 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight9/10

Deployed on 5 chains (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight10/10

Median 78 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight2/10

No public status page found. Nothing about its reliability can be verified from outside — including by us.

measured 2026-10-19 14:39 UTC
Gains Network full scorecard
5

Jupiter Perps

Status page, no readable feed
6.16100% measured

Latency 10/10. Median 117 ms to first byte from the live interface, our own probe, slowest sample discarded, single location.

Security audits 2/10. No published audits on DefiLlama.

Not for: Anyone putting size into it: it holds funds in smart contracts with no published audit. Fine to experiment with, not to trust with a position you cannot afford to lose.

Every criterion, with evidence
SCALEScale of value trusted
30% weight8/10

$745.0M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight2/10

No published audits on DefiLlama. Not proof of insecurity, but an unaudited contract holding funds is a red flag.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight8/10

Live 2.7 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight6/10

Deployed on 1 chain (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight10/10

Median 117 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight5/10

A status page exists but publishes no feed we could parse — readable, not auditable. Part of that may be our blind spot (we read three platforms). This is a company-wide status page shared with 2 other rated products, so it describes the company's systems rather than this product alone.

measured 2026-10-19 14:39 UTCevidence ↗
Jupiter Perps full scorecard
6

MUX Protocol

Not published
6.16100% measured

Track record 9/10. Live 4.0 yr (time since DefiLlama listing (a floor on true age)).

Status transparency 2/10. No public status page found.

Not for: Anyone who needs to know when it breaks: no status feed, so you find out it is down by discovering it yourself.

Every criterion, with evidence
SCALEScale of value trusted
30% weight4/10

$9.8M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight6/10

1 published audit with links (DefiLlama) — read the findings yourself. Audits reduce risk, they do not remove it.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight9/10

Live 4.0 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight9/10

Deployed on 5 chains (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight9/10

Median 231 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight2/10

No public status page found. Nothing about its reliability can be verified from outside — including by us.

measured 2026-10-19 14:39 UTC
MUX Protocol full scorecard
7

dYdX

Status page, no readable feed
5.84100% measured

Track record 10/10. Live 8.0 yr (since launch (no later than Oct 19, 2018), public record).

Security audits 2/10. No published audits on DefiLlama.

Not for: Anyone putting size into it: it holds funds in smart contracts with no published audit. Fine to experiment with, not to trust with a position you cannot afford to lose.

Every criterion, with evidence
SCALEScale of value trusted
30% weight6/10

$70.0M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight2/10

No published audits on DefiLlama. Not proof of insecurity, but an unaudited contract holding funds is a red flag.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight10/10

Live 8.0 yr (since launch (no later than Oct 19, 2018), public record). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight6/10

Deployed on 1 chain (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight9/10

Median 363 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight5/10

A status page exists but publishes no feed we could parse — readable, not auditable. Part of that may be our blind spot (we read three platforms).

measured 2026-10-19 14:39 UTCevidence ↗
dYdX full scorecard
8

Ostium

Not published
5.76100% measured

Latency 10/10. Median 195 ms to first byte from the live interface, our own probe, slowest sample discarded, single location.

Status transparency 2/10. No public status page found.

Not for: Anyone who needs to know when it breaks: no status feed, so you find out it is down by discovering it yourself.

Every criterion, with evidence
SCALEScale of value trusted
30% weight4/10

$4.1M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight6/10

1 published audit with links (DefiLlama) — read the findings yourself. Audits reduce risk, they do not remove it.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight8/10

Live 2.2 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight6/10

Deployed on 1 chain (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight10/10

Median 195 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight2/10

No public status page found. Nothing about its reliability can be verified from outside — including by us.

measured 2026-10-19 14:39 UTC
Ostium full scorecard
9

Drift

Not published
5.56100% measured

Status change: After the April 1, 2026 exploit the original Drift program is paused; the team is relaunching as Velocity.

Track record 9/10. Live 4.9 yr (time since DefiLlama listing (a floor on true age)).

Status transparency 2/10. No public status page found.

Not for: Anyone who wants a battle-tested venue. Little value sits here yet, so it has not been stress-tested at scale — early-adopter territory.

Every criterion, with evidence
SCALEScale of value trusted
30% weight3/10

$642,589 in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight6/10

1 published audit with links (DefiLlama) — read the findings yourself. Audits reduce risk, they do not remove it.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight9/10

Live 4.9 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight6/10

Deployed on 1 chain (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight9/10

Median 292 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight2/10

No public status page found. Nothing about its reliability can be verified from outside — including by us.

measured 2026-10-19 14:39 UTC
Drift full scorecard
10

ApeX Protocol

Not published
5.54100% measured

Multi-chain reach 9/10. Deployed on 5 chains (DefiLlama).

Security audits 2/10. No published audits on DefiLlama.

Not for: Anyone putting size into it: it holds funds in smart contracts with no published audit. Fine to experiment with, not to trust with a position you cannot afford to lose.

Every criterion, with evidence
SCALEScale of value trusted
30% weight6/10

$30.9M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight2/10

No published audits on DefiLlama. Not proof of insecurity, but an unaudited contract holding funds is a red flag.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight8/10

Live 2.3 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight9/10

Deployed on 5 chains (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight9/10

Median 333 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight2/10

No public status page found. Nothing about its reliability can be verified from outside — including by us.

measured 2026-10-19 14:39 UTC
ApeX Protocol full scorecard
11

Extended

Not published
5.46100% measured

Latency 10/10. Median 66 ms to first byte from the live interface, our own probe, slowest sample discarded, single location.

Security audits 2/10. No published audits on DefiLlama.

Not for: Anyone putting size into it: it holds funds in smart contracts with no published audit. Fine to experiment with, not to trust with a position you cannot afford to lose.

Every criterion, with evidence
SCALEScale of value trusted
30% weight7/10

$115.6M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight2/10

No published audits on DefiLlama. Not proof of insecurity, but an unaudited contract holding funds is a red flag.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight6/10

Live 1.6 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight8/10

Deployed on 2 chains (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight10/10

Median 66 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight2/10

No public status page found. Nothing about its reliability can be verified from outside — including by us.

measured 2026-10-19 14:39 UTC
Extended full scorecard
12

edgeX

Not published
5.26100% measured

Latency 10/10. Median 155 ms to first byte from the live interface, our own probe, slowest sample discarded, single location.

Security audits 2/10. No published audits on DefiLlama.

Not for: Anyone putting size into it: it holds funds in smart contracts with no published audit. Fine to experiment with, not to trust with a position you cannot afford to lose.

Every criterion, with evidence
SCALEScale of value trusted
30% weight6/10

$60.0M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight2/10

No published audits on DefiLlama. Not proof of insecurity, but an unaudited contract holding funds is a red flag.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight6/10

Live 1.1 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight9/10

Deployed on 5 chains (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight10/10

Median 155 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight2/10

No public status page found. Nothing about its reliability can be verified from outside — including by us.

measured 2026-10-19 14:39 UTC
edgeX full scorecard
13

Paradex

Status page, no readable feed
5.2100% measured

Latency 10/10. Median 67 ms to first byte from the live interface, our own probe, slowest sample discarded, single location.

Security audits 2/10. No published audits on DefiLlama.

Not for: Anyone putting size into it: it holds funds in smart contracts with no published audit. Fine to experiment with, not to trust with a position you cannot afford to lose.

Every criterion, with evidence
SCALEScale of value trusted
30% weight6/10

$30.1M in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight2/10

No published audits on DefiLlama. Not proof of insecurity, but an unaudited contract holding funds is a red flag.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight6/10

Live 1.1 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight6/10

Deployed on 1 chain (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight10/10

Median 67 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight5/10

A status page exists but publishes no feed we could parse — readable, not auditable. Part of that may be our blind spot (we read three platforms).

measured 2026-10-19 14:39 UTCevidence ↗
Paradex full scorecard
14

Synthetix

Status page, no readable feed
4.86100% measured

Status change: Synthetix's current product is an order-book exchange on Ethereum (since December 2025); DefiLlama's listings may not yet reflect it.

Latency 10/10. Median 52 ms to first byte from the live interface, our own probe, slowest sample discarded, single location.

Security audits 2/10. No published audits on DefiLlama.

Not for: Anyone putting size into it: it holds funds in smart contracts with no published audit. Fine to experiment with, not to trust with a position you cannot afford to lose.

Every criterion, with evidence
SCALEScale of value trusted
30% weight3/10

$883,499 in value locked (DefiLlama). Capital is the market voting with its money — scored on a log scale, so the jump from $10M to $100M counts for more than $10B to $20B.

measured 2026-10-19 14:40 UTCevidence ↗
AUDITSecurity audits
26% weight2/10

No published audits on DefiLlama. Not proof of insecurity, but an unaudited contract holding funds is a red flag.

measured 2026-10-19 14:40 UTCevidence ↗
TRACKTrack record
18% weight8/10

Live 2.5 yr (time since DefiLlama listing (a floor on true age)). Most failures happen young, so time survived is real signal.

measured 2026-10-19 14:40 UTCevidence ↗
REACHMulti-chain reach
10% weight8/10

Deployed on 3 chains (DefiLlama). Breadth is a minor tie-breaker — a focused single-chain protocol can still be excellent.

measured 2026-10-19 14:40 UTCevidence ↗
PERFLatency
8% weight10/10

Median 52 ms to first byte from the live interface, our own probe, slowest sample discarded, single location. Compared only within this category.

measured 2026-10-19 14:39 UTCevidence ↗
STATUSStatus transparency
8% weight5/10

A status page exists but publishes no feed we could parse — readable, not auditable. Part of that may be our blind spot (we read three platforms).

measured 2026-10-19 14:39 UTCevidence ↗
Synthetix full scorecard

How perpetual dexs are scored

Perp venues need deep liquidity backing the book and audited engines.

Scale of value trusted
30%
Security audits
26%
Track record
18%
Multi-chain reach
10%
Latency
8%
Status transparency
8%

Not scored for perpetual dexs: security record, peg stability. Latency is timed at the live interface for every service here, so the comparison is like for like. Full methodology

Head to head

Checked against the disqualifiers

These services have an incident that could trigger an absolute disqualifier. Each was checked against the evidence and kept in the ranking; the reasons are below.

The April 1, 2026 exploit (about $295M) is recorded in DefiLlama with no returned amount. Drift published a recovery plan on May 5, 2026: affected users receive claim tokens for their verified losses, funded by a pool with up to $127.5M committed by Tether plus partner and protocol-revenue contributions. Repayment is incomplete and may take years, but users have not been left without restitution, so the rule does not apply. The hack still counts in the security record wherever that criterion is weighted.

CoinDesk: Drift outlines a recovery plan ↗DL News: Drift to issue recovery tokens ↗

The July 15, 2026 oracle-key exploit (about $23.75M) is recorded in DefiLlama with no returned amount. Ostium's compensation plan fully reimburses 3,321 of 3,666 affected wallets (90.6%) in its first phase; larger losses go to a second phase opening October 30, 2026. Restitution is in progress, so the rule does not apply.

Crypto Briefing: Ostium recovery plan repays 3,321 wallets ↗KuCoin News: Ostium second-stage compensation plan ↗

Perpetual DEXs: questions about this ranking

What is the best perp DEX on this ranking?+

On October 19, 2026 data, GMX scores highest at 7.04/10, ahead of Aevo (6.68) and Hyperliquid (6.6). "Best" means best-evidenced on value locked, audits, track record, reach, latency and status transparency. It does not cover fees, funding, liquidity or hack history.

Why does Hyperliquid rank third with the most value locked?+

DefiLlama lists no published audit for it, which scores 2 on a criterion worth 26%. Its $6.78B value locked scores 9, not 10, because the scale is logarithmic.

Why is hack history not part of the score?+

The category weights put 0% on security record. The incident table on this page lists what DefiLlama records so you can weigh it yourself. Separately, Drift and Ostium were checked against the “funds never returned” disqualifier and kept, because restitution plans are in progress.

Is a high score a sign that a venue is safe?+

No. It means the venue is stronger on public, checkable signals as of the collection date. Leveraged trading can lose more than you expect even on a well-run venue.

How old is the data?+

All scores are from the October 19, 2026 collection, with each criterion time-stamped on the rating card. Value locked moves daily.

Do venues pay to be listed?+

No. There are no affiliate links, paid placements or sponsored entries. See how CoinPulse is funded. *Provenance:* scores, ranks, metrics and timestamps are copied from the CoinPulse collection of October 19, 2026 (14:39 UTC). The incident table is from DefiLlama's public hack database, checked October 23, 2026, and is context, not part of the score. *Update rule:* ranks and use-case lines regenerate with each collection; the incident table should be re-checked at the same time.

Last collection 2026-10-19: 254 registered services checked, 234 answered the latency probe; status feeds read over a 90-day window.