Exchange Trust Score

An exchange can prove its own reserves. Only an independent arbiter can rate all three side by side. Every input below is a published fact from the exchange's own proof-of-reserves page — scored by an open, deterministic formula.

Snapshot date: 2026-07-22. Scores recompute automatically on every new snapshot.

Gate.io

source: gate.com/proof-of-reserves · report 2026-06-18

88/ 95

Total reserves $8.18B · 115% · excess +$1.11B

Reserve coverage115% → 40/40

Report recency85d ago → 18/25

Method: Merkle tree + zk-SNARKs20/20

Public walletsyes → 10/10

Platform-wide total reserve ratio as published; reliable per-coin ratios pending our next snapshot. Per-coin exchange wallet balances shown against customer balances; audit code open-sourced on GitHub.

Claims first mainstream platform committed to 100% reserves. · Excess reserve published explicitly: +$1.11B over customer balances.

OKX

source: okx.com/proof-of-reserves · report 2026-07-07

78/ 95

PoR report #42 · $26.2B reserves

Reserve coverage101% → 30/40

Report recency66d ago → 18/25

Method: zk-STARK proofs, Merkle tree20/20

Public walletsyes → 10/10

Lowest of the eight published major-asset ratios (USDC 101%); BTC 105%, ETH 103%, USDT 112%. On-chain wallet addresses published and verifiable.

42 consecutive reports — the longest public PoR streak of the three. · Per-account inclusion proof available to every user (Merkle path).

Bybit

source: bybit.com/proof-of-reserves

Bybit's PoR page requires an interactive browser step our snapshot harness is being extended to handle. The score publishes here the moment real data lands — we never score from guesses. What we can already verify: licensed entities in Vienna, Cyprus and Dubai; ~$13.1B daily volume self-reported.

MEXC

source: mexc.com/proof-of-reserves

MEXC publishes proof-of-reserves pages, but they sit behind the same datacenter block as its fee schedule — snapshot pending via our browser harness. The score publishes the moment real data lands; until then this card stays honestly empty. What we can already verify: the public spot and futures APIs respond from our infrastructure directly.

The formula (open, versioned in git)

Coverage /40: ≥110% → 40 · ≥105% → 36 · ≥100% → 30 · ≥95% → 12 · below → 0. We use the weakest reliably-published ratio, not the flattering one.

Recency /25: report ≤45 days old → 25 · ≤90d → 18 · ≤180d → 10 · older → 4.

Verification method /20: zero-knowledge proofs → 20 · third-party auditor → 14 · self-attested → 6.

Public wallets /10: verifiable on-chain addresses or balances → 10, otherwise 0.

An incident-history penalty is reserved for when we hold enough documented history to apply it fairly to all three. Until then the maximum is 95, not 100 — the missing five points are the honesty margin.

And our own proof? We are not a custodian — there are no reserves to prove. What we prove instead is our track: every paper deal lands in an append-only ledger, losing streaks included, on the virtual floor. Same standard we hold the exchanges to: published data or it did not happen.

Proof-of-reserves shows assets, not liabilities beyond the published customer balances, and is a point-in-time attestation — it is a transparency signal, not a solvency guarantee. All figures belong to their exchanges and carry the snapshot date above. This page is information, not investment advice. See also: fee comparison · bot types truth · API hub.