Crypto exchange API latency by city

Measured 2026-11-12, repeated monthly

Binance, Bybit, OKX and Hyperliquid answer fastest from Tokyo: under 10 ms to Binance. Coinbase is closest from Chicago (about 30 ms), Kraken from London (about 15 ms). From Europe the Asian exchanges are about 215 to 260 ms away. We measured from 21 cities where you can rent a server by the hour.

For a trading bot that races other traders, the city matters more than the size of the server. Below is the round trip from each city to six exchanges, so you can pick the city before you deploy.

Best city for each exchange

ExchangeClosest cityRound trip
BinanceTokyounder 10 ms
BybitTokyounder 10 ms
OKXTokyo25 ms
CoinbaseChicago30 ms
KrakenLondon15 ms
HyperliquidTokyo15 ms

Round trip from every city, ms

Rounded to 5 ms. Binance and Bybit do not serve addresses in the United States, so US cities show no number for them.

Asia-Pacific

CityBinanceBybitOKXCoinbaseKrakenHyperliquid
Tokyo<10<10~25~185~250~15
Singapore~75~75~95~255~185~85
Sydney~105~105~115~225~290~125
Mumbai~140~140~140~215~155~155
Bangalore~155~155~125~220~190~165

Europe

CityBinanceBybitOKXCoinbaseKrakenHyperliquid
London~215~215~255~120~15~225
Amsterdam~235~235~255~120~35~250
Nuremberg~230~235~250~120~40~250
Frankfurt~230~230~255~110~45~240
Stockholm~245~250~255~130~45~255
Helsinki~260~260~255~135~50~260

North America

CityBinanceBybitOKXCoinbaseKrakenHyperliquid
Chicagono US accessno US access~150~30~110~140
Atlantano US accessno US access~165~50~105~170
New Jerseyno US accessno US access~170~50~95~160
Miamino US accessno US access~180~65~130~190
Los Angelesno US accessno US access~115~70~155~115
Silicon Valleyno US accessno US access~120~85~155~125
Seattleno US accessno US access~110~100~170~100
Dallasno US accessno US access~145~110~140~165
Toronto~140~140~155~45~105~150

South America

CityBinanceBybitOKXCoinbaseKrakenHyperliquid
São Paulo~325~325~280~135~180~330

Exact numbers: exchange-latency.csv (CC BY 4.0, cite NoctHost). The same data with the measurement script is on GitHub and archived on Zenodo, DOI 10.5281/zenodo.24247822.

What the numbers say

  • Binance, Bybit, OKX and Hyperliquid are closest from Tokyo. Singapore is the next best, at 75 to 95 ms.
  • Coinbase is closest from the eastern US: Chicago, then New Jersey and Atlanta.
  • Kraken is closest from Europe: London first, then Amsterdam, Nuremberg and Frankfurt.
  • A bot that trades on Binance and Coinbase at once has no single best city. Run one server near each exchange, or pick by the exchange where speed matters most.

How we measured

In each city we started the smallest server (1 vCPU, 1 GB), sent 11 requests over one connection to the exchange's public server-time method and took the median of requests 2 to 11. That is the time a request takes to reach the exchange and come back, without DNS, TCP and TLS setup. We ran it twice a few minutes apart and kept the lower number; the runs differed by under 20 ms.

The public time method shows where an exchange serves its API. Order gateways and WebSocket feeds may run elsewhere, so check from your own server with the endpoint your bot uses. The same measurement for Binance:

URL=https://api.binance.com/api/v3/time
for i in $(seq 11); do set -- "$@" -o /dev/null "$URL"; done
curl -s -w '%{time_starttransfer} %{time_pretransfer}\n' "$@" \
  | tail -n +2 | awk '{print ($1-$2)*1000}' | sort -n | sed -n 5p

Frequently asked questions

Where should I host a Binance trading bot?

Tokyo. Binance's API answered in under 10 ms from Tokyo, about 75 ms from Singapore and over 200 ms from Europe. Binance does not serve addresses in the United States.

Where should I host a Coinbase or Kraken bot?

Coinbase: Chicago (about 30 ms), with New Jersey and Atlanta close behind. Kraken: London (about 15 ms), or Amsterdam, Nuremberg and Frankfurt at 30 to 45 ms.

Does latency matter for my bot?

Only for bots that race other traders for fills: market making, arbitrage, liquidations. A bot that trades on candle closes, timers or signals sees no difference between 10 and 250 ms.

What exactly was measured?

From a fresh server in each city: 11 requests to the exchange's public server-time method over one connection, median of requests 2 to 11. That is the request round trip to the exchange, without DNS, TCP and TLS setup. Two runs a few minutes apart; the runs differed by under 20 ms.

Can I check it myself?

Yes. Rent the smallest server in two or three candidate cities for an hour, run the command shown on this page, keep the fastest city and destroy the rest. An hour of testing costs a few cents.

Run your bot next to the exchange

Servers in all 21 cities, billed by the hour. See VPS for trading bots and picking the region for a trading bot.

Get started