What we measure

Hyperliquid: WebSocket round-trip to api.hyperliquid.xyz via AWS CloudFront, plus direct TCP latency to Hyperliquid validator nodes in Tokyo (AZ1, AZ2, AZ4) and real order-to-fill timing.

Solana: QUIC handshake latency to every voting Solana mainnet validator (~760 nodes), real-time leader-rotation tracking, stake-weighted best-co-location rankings, and Jito Block Engine paths across all 8 regional endpoints (Amsterdam, Frankfurt, London, Dublin, New York, Salt Lake City, Tokyo, Singapore).

SUI: TCP+TLS handshake latency to every active SUI mainnet validator with Mysticeti consensus. Reports time-to-quorum (latency to accumulate signatures from 2/3 of stake-weighted voting power) and time-to-90% supermajority — the metrics that actually bound transaction finality on a DAG-BFT chain.

Centralized exchanges (CEX): WebSocket and FIX latency to Binance, Coinbase, Kraken, OKX, Bybit, Bitmex, Deribit across spot, perpetuals, expiry futures, and options venues. Cross-exchange comparison mode picks the closest venue for a given product from any probe.

Prediction markets: REST round-trip to Polymarket (clob.polymarket.com, origin AWS eu-west-2 London) and Kalshi (api.elections.kalshi.com, origin AWS us-east-2 Ohio).

Arbitrum One: Transaction submission round-trip to the Arbitrum One sequencer (arb1-sequencer.arbitrum.io, AWS us-east-2 Ohio) — the single machine that orders every transaction on the chain. Co-location guide for Timeboost express-lane bidders and first-come-first-served flow.

Robinhood Chain: Transaction submission round-trip to the Robinhood Chain sequencer (sequencer.mainnet.chain.robinhood.com, AWS us-east-2 Ohio) — Robinhood's Arbitrum-stack L2 for tokenized stocks and real-world assets. Ordering is strictly first-come-first-served with no fee-based priority, so distance to the sequencer is queue position.

Base: Transaction submission round-trip to the Base sequencer (mainnet-sequencer.base.org) — the single endpoint every Base transaction must reach, since the OP Stack has no public mempool. Ordering combines priority fee and arrival time, and a later-arriving transaction cannot buy its way into an earlier Flashblock, so network distance decides which 200ms Flashblock a transaction lands in.

X Layer: Transaction submission round-trip to X Layer, OKX's OP Stack Layer 2 (rpc.xlayer.tech) — the OP Stack has no public mempool, so every transaction must reach the sequencer to be included. Blocks are produced roughly every second, and arrival time decides which block a transaction lands in.

Ink: Transaction submission round-trip to Ink, Kraken's OP Stack Layer 2 (rpc-gel.inkonchain.com) — the same host the Superchain registry lists as the sequencer RPC. Uniquely among the chains here it sits behind no CDN, so connection timings are real probe-to-sequencer distance. Blocks are produced roughly every second.

OP Mainnet: Transaction submission round-trip to Optimism's own Layer 2 and the original OP Stack chain, via its dedicated sequencer endpoint (mainnet-sequencer.optimism.io). Measurements place the sequencer in Europe. Blocks are produced every two seconds, and arrival time decides which block a transaction lands in.

ZKsync Era: Transaction submission round-trip to Matter Labs' zero-knowledge rollup (mainnet.era.zksync.io), sequenced by a single operator. Measurements place it in North America — the counterexample to Ink and OP Mainnet, which both answer fastest from Europe.

HFT oracle gateways: Pyth Pro/Lazer (Tokyo bare metal), Switchboard Crossbar (GCP eu-west4), Chainlink Data Streams (Cloudflare US). Co-location guide for trading on oracle-dependent prices.

Why infrastructure location matters

For trading firms racing to be first into a price move, physical distance to exchange matching engines and blockchain validators dominates total latency. A server in the wrong city can lose by tens to hundreds of milliseconds — enough to lose every arbitrage and MEV opportunity to better-located competitors. This site publishes the actual measured numbers, continuously, so operators can make data-driven hosting decisions instead of guessing.

CEX Hyperliquid Prediction Markets Solana SUI Arbitrum One Robinhood Chain Base X Layer Ink OP Mainnet ZKsync Era Oracle Arb Routes API Glassnode Studio→

← Back to map

About OP Mainnet Latency

Real-time latency monitoring for OP Mainnet.

What is OP Mainnet Latency?

OP Mainnet is Optimism's own Layer 2 and the original OP Stack chain. Every transaction is ordered by a single sequencer, and there is no other route onto the chain — whatever path a transaction takes, it has to reach that sequencer to be included.

How To Read These Numbers

OP Mainnet publishes a dedicated sequencer endpoint, so these measurements follow the same path a transaction takes rather than a general-purpose node that forwards it on.

That endpoint sits behind a global load balancer, which accepts the connection close to wherever you are and carries the request onward from there. The round trip we report is the real journey to the sequencer and back, but connection-setup timings would only measure the distance to that load balancer, so they are not meaningful here and we do not show them.

Our measurements put the sequencer in Europe — European probes answer fastest and probes in Ohio are as far from it as they are from a known European endpoint. We stop short of naming a country, because a load balancer hides the origin and the evidence only narrows it to a continent.

Why It Matters

OP Mainnet produces a block every two seconds. Fees compete for position inside a block, but arrival time decides which block a transaction makes at all — and once that block is built, a later transaction cannot buy its way into it. Paying for priority raises the price of speed rather than removing the need for it. That distance is what this map measures.

Methodology

Target	mainnet-sequencer.optimism.io
Measurement	Each probe submits a fixed signed transaction via eth_sendRawTransaction.
Metric	Time to first byte of the response over a warm connection — network round trip plus a few milliseconds of validation. p50 over 60-second windows.

Research Project Disclaimer

OP Mainnet Latency is a research project provided for informational and educational purposes only. The latency metrics displayed are measured from specific geographic probe locations using our distinct server environments. Actual latency may differ. Real-world network performance depends on numerous factors including ISP routing, local network congestion, and hardware. These numbers should be viewed as estimates and directional baselines. This project is not affiliated with or endorsed by OP Mainnet or Optimism.

Want this data programmatically?

A REST API for all latency data is in the works. Join the waitlist and we'll email you once at launch →

Glassnode Latency Monitor — Real-Time Crypto Trading Infrastructure Latency