- p50
- 170.0ms
- p95
- 213.0ms
- p99
- 238.0ms
2,661 samples
Every number traces one Hyperliquid market event from the exchange timestamp to a decoded, validated order book. Measured live from five regions, starting with Tokyo, where most of the Hyperliquid validator set runs.
Time from the Hyperliquid block timestamp to a decoded, validated, uncrossed top-of-book quote. Quicknode Hypersonic is pre-execution Optimistic BBO over SUB bbo; the other sources are post-execution books. Measured by the AWS runner in Tokyo over the latest rolling five-minute window.
2,661 samples
2,661 samples
2,661 samples
2,948 samples. Predicted from the block proposal before execution, then confirmed once the block executes. The other sources are executed books. Measured from an AWS client in Tokyo, separate from the selected runner, so it’s shown for reference and not ranked.
Market opens and closes are when trading conditions change fastest. Every session is drawn on the chart below, so you can see how each source performs through each handoff.
Brazil, Canada, United States
19:55 UTC, in 1h 21m
Completed Nov 14, 08:30 to 08:50 UTC
Each point is the rolling five-minute value at that time, in UTC. Purple windows mark each market open, from 15 minutes before to 45 minutes after, including the 13:30 UTC US open. Grey windows mark Hyperliquid network maintenance and upgrades from its status page. The bars above the lines show which APAC, EMEA, and Americas markets were trading.
Every market event is timed on every source at once. This is the share of those races each source won outright over the last hour.
30,569 head-to-head comparisons across 119 intervals. 0.0% ended in a tie.
A ping tells you how fast a server answers. A trading system cares how long a market event takes to become usable data. This benchmark times exactly that, for every event, on every source at once.
The clock starts at the timestamp Hyperliquid stamps on the event, not when a request was sent, so network and processing time both count.
The clock stops once the event is decoded, validated, and applied to a canonical order book on the runner. Raw bytes arriving early do not count.
Each event is matched across sources, so every comparison is a head-to-head race on identical data rather than two separate samples.
Runners sync with chrony against a 5ms maximum offset and publish their live clock error bound with every measurement.
How the Hyperliquid benchmark works and how to read the results.
Questions about these numbers, Hypersonic, or sizing a setup for your region? Send a note and the Hyperliquid team will follow up.
NEW: Live Hyperliquid latency benchmarkEvery market event timed from exchange to order book, in 5 regions.
Compare live latency