Gate.io

A user-fund incident from July 2026 is still open and users are still uncompensated as we publish this. Every other incident in our tables is history to be judged on how it was resolved; this one has not been resolved, which makes it a live fact about the venue rather than a record of one.

The reserve proof is technically decent and structurally incomplete. Gate publishes a Merkle-and-zk proof showing a 115% reserve ratio with open-source verification, which is more cryptographic effort than most of the category makes. What it lacks is a periodic third-party balance attestation. The external sign-off it displays is a review of the verification code — confirmation that the tool works, not confirmation that the money is there. Those are different claims and the distinction is the entire value of an attestation.

The protection story is weaker still. The widely quoted "$200m SAFU" has no published proof behind it. The fund that does exist is a per-contract derivatives backstop with no aggregate figure and no on-chain address, and Gate’s own documentation states it does not cover individual trading losses.

We score custody 4 out of 10. The reason it is not lower is the genuine cryptographic work in the proof; the reason it is not higher is that an open incident outranks a good Merkle tree.

We measured 3,028 ms of API response latency — just over three seconds, and roughly six times the next slowest venue in the run. Crypto.com came second-worst at 572 ms; Coinbase answered in 183 ms.

For someone placing an occasional manual order this is an irritation rather than a risk. For anything automated it is disqualifying, and it is also a signal about infrastructure investment that is worth reading alongside everything in the section above.

Cost and depth are unremarkable: 0.2063% on a $10,000 bitcoin buy, rising to 0.2206% at $100,000, on 10.5 BTC of depth within 0.1% of best ask — among the thinnest books we priced.

What Gate.io has is breadth, scoring 9 out of 10 on assets with one of the widest listings anywhere. As with KuCoin, that is a reason to execute here and not a reason to hold here, and with an open incident on the record the distinction is sharper than usual.