Which exchange stays up when the market moves
We probe 10 venues every 15 seconds from a fixed vantage point and publish every measurement — including the numbers exchanges never report about themselves.
Live board
Ranked by availability, then by median response time.
| Exchange | Availability | Uptime | Median | p95 | Clock offset | Data age | Status |
|---|---|---|---|---|---|---|---|
|
1
|
|
100% | 4 ms | 15 ms | −2 ms | 10 ms | Operational |
|
2
|
|
100% | 6 ms | 144 ms | −2 ms | ≈0 ms | Operational |
3
KuCoin
Global
|
|
99.85% | 14 ms | 38 ms | ±0 ms | 481 ms | Degraded |
|
4
|
|
99.85% | 16 ms | 43 ms | ±0 ms | ≈0 ms | Degraded |
|
5
|
|
99.85% | 19 ms | 129 ms | +1 ms | ≈0 ms | Degraded |
|
6
|
|
99.85% | 73 ms | 208 ms | −2 ms | ≈0 ms | Degraded |
|
7
|
|
99.85% | 173 ms | 272 ms | −2 ms | 1,131 ms | Degraded |
|
8
|
|
99.85% | 256 ms | 715 ms | +3 ms | ≈0 ms | Degraded |
9
MEXC
Global
|
|
98.8% | 28 ms | 83 ms | +1 ms | 427 ms | Failing |
10
HTX
Global
|
|
98.0% | 12 ms | 35 ms | ±0 ms | ≈0 ms | Failing |
What nobody else publishes
Three numbers you cannot get from an exchange's own status page.
Clock offset
How far a venue's server clock sits from the consensus of all venues measured. Your order timestamps, funding settlement and API rate windows are all stamped by that clock.
All 10 venues sit within 3 ms of consensus. No venue is measurably drifting.
Market-data age
How old a quote already is at the instant the venue hands it to you, in the venue's own clock. Stale quotes are the gap between the price you see and the price you get.
Coinbase — stalest · OKX ≈0 ms — freshest
Uptime under stress
Availability measured only inside the busiest decile of 5-minute windows in the record — currently those where the consensus BTC price moved at least 0.18%. The threshold rescales with the market rather than being fixed, because a fixed one either never fires in a calm week or waves through noise in a wild one. Uptime averaged over a quiet market is decoration.
HTX — 519 probes
Recent incidents
Opened after 2 consecutive failed probes; closed after 2 consecutive good ones.