Solana RPC latency checker

This Solana RPC latency checker sends a few read calls to an endpoint you paste in and reports how long each took and how far the endpoint trails the public cluster. The calls come from our server, so the numbers are server-to-RPC latency from Vercel, not what your own machine or your users see.

No sign-up. Runs in your browser. Last reviewed November 4, 2026 by the Builderz team.

How it works

What is measured, and from where

Our server makes the calls, because this page cannot reach RPC endpoints from your browser. Each call is timed on the server with a monotonic clock, and the time includes DNS resolution and TLS setup from our server, so a fast endpoint reads a little slower than a warm client would see. The region shown in the result is the Vercel region that ran the check. A different region, or your own machine, can give different numbers.

The four calls

getHealth asks the node whether it considers itself caught up. getSlot, at processed commitment, returns the newest slot the node has seen. getLatestBlockhash is what a client needs before it can build a transaction. getVersion returns the node software version. The calls run one after another, not in parallel, so they do not distort each other.

Slot lag

After the four calls we read getSlot once from the public mainnet-beta cluster and subtract the endpoint slot. A positive number means the endpoint trails. The two reads happen seconds apart and slots advance fast, so treat the lag as approximate. A few slots of difference is normal noise at this resolution.

What is not measured

Websocket subscriptions, sendTransaction and how often it lands, rate limits under load, and the geographic latency between your users and the provider are all outside this check. One run is also a single sample. Run it several times, and compare with a second provider.

What happens to your URL

The URL goes to our server for the measurement, and only to vetted public https hosts: no redirects, no private addresses, an 8 second limit per call. It travels in the request body, not the query string, so it stays out of edge request logs, and our code does not store or log it. The result shows the URL with the query string removed and any key-like path segment masked. If your URL has a key in it, you can rotate the key afterwards.

Questions about this tool

Why is this the latency from your server and not from my machine?

Our content security policy stops this page from calling other hosts from your browser, so the server makes the calls. The number tells you how the endpoint responds to a Vercel function. For your own latency, time the same calls from your app or your hosting region.

What do fast, ok, and slow mean here?

They are our own working thresholds: under 150 ms is fast, under 400 ms is ok, and above that is slow. They are not a standard, and a number that suits a trading bot may not suit a dashboard.

Is it a problem if my endpoint is a few slots behind?

Not necessarily. The two reads are seconds apart and the public cluster is itself a busy endpoint. A lag that stays high across several runs is more telling than one reading.

Does it work with a private or local endpoint?

No. The endpoint must be reachable from the public internet over https. Local, private-network, and plain http URLs are rejected.