About MPATEL Speed Test

MPATEL Speed Test is a free set of internet tests that run in your browser: speed, ping, DNS, traceroute and IP checks. This page explains who runs it, how each test measures, which outside servers take part and what a browser simply cannot measure.

Who runs this site

MPATEL Speed Test is run by MPATEL. It is an independent site: no internet provider, app maker or game publisher pays for it or checks its results. The line at the foot of every page shows who maintains the site and when that page last changed.

There are no accounts, no adverts and no analytics scripts. Our own server sends you the pages and, only when Cloudflare cannot tell your browser its address, answers one small address lookup. The tests run in your browser and talk directly to the measurement servers named on this page.

The explanations on this site were drafted with the help of an AI writing assistant. Every figure that is not our own measurement comes from a source we link next to it, such as a regulator, an app maker’s help page or a technical standard.

Why it exists

It started with a home broadband line that kept dropping, slowing down and disconnecting. A single speed test at a good moment proved nothing, so we wrote a monitor that measured the line all day and kept a record that could be shown to the provider.

That monitor is now the internet stability test. The other tools grew around it, so that you can answer the next questions yourself: is it the Wi-Fi or the line, is it DNS, and where on the route does the delay start?

How each test works

Each section below belongs to one page of the site. The “How this test works” link at the bottom of that page leads here. Where a test uses mobile data, the figure comes from our own runs; a slower line uses less.

Internet Speed Test

The internet speed test downloads from and uploads to Cloudflare’s nearest speed test server, and the result names that server, for example “Cloudflare, New Delhi (DEL)”. It first times about 20 small HTTPS requests with the line quiet, then tests parallel connections, adding one every half second up to the chosen mode’s limit (two or four). A flat reading does not stop this warm-up early. The result averages throughput after the last connection warms up; a stable result uses the final three steady seconds. Short, data-capped tests use a shorter window and are labelled as limited. Ping is timed every 300 ms during both transfers to show latency under load.

Each size has a data cap: Data saver can use up to 37 MB, Standard up to 217 MB and Gigabit up to 876 MB. In our own run on a line of about 200 Mbps, Data saver used 35 MB and Standard 210 MB in about 14 seconds. Data saver stops at its cap before the speed settles on lines faster than about 60 Mbps download (25 Mbps upload), and Standard above about 200 Mbps. The result then shows as “at least” that speed, so fibre users should pick Gigabit. The optional second test on M-Lab runs only after you agree, because M-Lab publishes your IP address.

M-Lab speed test (the test in Google Search)

The M-Lab speed test page runs NDT7, the Measurement Lab test that also appears in Google Search. It asks M-Lab’s Locate service for a nearby server, then opens one encrypted WebSocket connection for download and one for upload, each for up to 10 seconds. Speed is all data moved divided by the time, which is how M-Lab counts it, and latency and TCP retransmissions come from the server’s own TCP statistics.

M-Lab publishes every result with your IP address as open data, so the test waits for you to tick the consent box each time. A full run moves whatever your line carries in up to 20 seconds; the Data saver run used 34.7 MB in our test. The optional Cloudflare run on the same page can use up to 37 MB more.

Broadband Speed Test

The broadband speed test uses the same Cloudflare test and the same data caps as the home page, then divides your result by the plan speed you type in to show the share of your plan you received. Latency and jitter are set next to TRAI’s fixed-broadband benchmarks as a guide only, because TRAI’s figures come from providers’ own measurements made another way.

Your plan, the time-of-day log and the complaint summary are kept in your browser. A test over Wi-Fi lowers the share of plan, so a low figure on Wi-Fi is not proof of a fault in the line.

Jio Speed Test

The Jio speed test page reads the network number (ASN) of your public address from Cloudflare and says whether it belongs to Reliance Jio (AS55836). A VPN, a private relay or an office network shows its own ASN instead, so the check can be wrong there. The speed test itself is the Cloudflare test described above, with the same caps.

The route trace does not start on your phone or router. It runs on a Globalping volunteer probe on Jio’s network, and the result prints that probe’s city, network and ASN. If no Jio probe is online, the page says so and uses a home or mobile connection elsewhere in India. The page is independent of Jio.

Airtel Speed Test

The Airtel speed test page tells Airtel’s mobile network (AS45609) apart from its home broadband network (AS24560) by the ASN Cloudflare reports for your address. After the Cloudflare speed test, you can trace a route from a Globalping probe on your own Airtel network, or on either of those two, with the probe named in the result.

Airtel runs its own speed test on its own servers, and the page links to it, so the two can show different numbers. When Globalping has no probe on the chosen Airtel network, the page falls back to another connection in India and tells you. The page is not made by Airtel.

WiFi Speed Test

The Wi-Fi speed test runs the Cloudflare test once per room and labels each run with the room name you type. A chart compares the rooms from the results stored in your browser. The test measures the whole path from your device to the internet, not the router alone.

A web page cannot see your Wi-Fi band, channel or signal strength, so the room name and the connection type you pick are what you tell it. To separate Wi-Fi from the line, the page suggests one run on a LAN cable.

5G and 4G Mobile Speed Test

The mobile speed test lets you choose a total cap of 10, 25 or 50 MB. About 70% of the cap goes to download and 30% to upload, over one or two connections for 3 to 8 seconds each way, and a live counter shows the megabytes used. The test stops at the cap, so a fast 5G line often shows “at least” a speed.

Your browser cannot tell 5G from 4G; Chrome reports “4g” for any fast connection. Check the network icon in your phone’s status bar while the test runs.

DNS Speed Test

The DNS speed test sends DNS-over-HTTPS (DoH) lookups from your browser to each public resolver in turn, never two at once. After one warm-up lookup that is thrown away, each round asks one popular name that resolvers usually have stored and one random name under mpatel.org that none can have stored. A lookup with no answer within 3 seconds is a failure, and the ranking uses our own score: 60% of the cached median, 40% of the uncached median and a penalty for failures.

This is DoH from the browser, not the plain DNS on port 53 that your phone or router normally uses, so treat the winner as a strong hint. Some resolvers, such as AdGuard and NextDNS, refuse requests from web pages and cannot be tested; the page lists them with the reason. The whole benchmark used about 12 KB in our run.

DNS Lookup and DNS Trace

The DNS lookup asks two or three public resolvers you choose, over DoH, and shows their answers side by side with TTLs, the CNAME chain and whether the resolver validated DNSSEC. Answers from public resolvers can differ from those of your provider’s resolver.

The delegation trace, from the root servers down to the domain, runs on a Globalping probe named in the result, because a browser cannot send raw DNS queries. The country check asks probes in five countries. Globalping publishes its measurements, so the domain you check becomes public.

Ping Test

The ping test sends one small HTTPS request per second to each server you tick and draws the times on a live graph. The first request to each server opens the connection and is not counted, a request with no answer in 4 seconds counts as failed, and the test stops after 10 minutes. For Cloudflare the browser’s timing data lets us remove the server’s processing time; the other servers include a little overhead.

These are HTTP round-trip times, not ICMP ping, and a failed request is not the same as a lost packet. The optional real ping sends 16 ICMP packets from a Globalping probe on your own network, or from another connection in India when your network has no probe, and names that probe. One minute with five servers moved about 0.3 MB in our run.

Gaming Ping Test

The gaming ping test times Amazon Web Services data centres in 11 regions, from Mumbai and Hyderabad to Singapore, Europe and the USA, one region at a time. Each region gets a warm-up request and five timed requests to its public DynamoDB address; the board shows the median and the spread between the 25th and 75th percentile as jitter.

We time cloud regions, not game servers, and never connect to a game. The verdict (great under 50 ms, good under 80, fair under 120) uses our own thresholds, because the game publishers do not publish a required ping. A full run took 66 requests and well under 0.1 MB.

Bufferbloat Test

The bufferbloat test measures ping to Cloudflare three times: with the line quiet, while it downloads, and while it uploads, with a timing request every 300 ms during each load. The rise is the loaded median minus the quiet median, shown as a box plot.

The A to F grade is our own scale for the larger rise: A under 30 ms, B under 60, C under 150, D under 400, otherwise F. Data saver stops at 25 MB down and 10 MB up; Standard at 150 MB and 60 MB. A short load may not fill your router’s queue, so a good Data saver grade on fast fibre is less certain.

Video Call Test

The video call test starts your camera and microphone only when you press its button. The preview and the sound level meter run inside your browser; nothing is recorded or sent anywhere, and both turn off when you press Stop or leave the page.

The network part is the Data saver speed test (up to 25 MB down and 10 MB up) with ping timed during both transfers. The verdicts compare your numbers with the figures Zoom, Google and Microsoft publish for their apps. We do not connect to their servers, so a real call can still behave differently.

Traceroute Online

The online traceroute runs on a Globalping probe, a small volunteer-run server in the city or network you choose, and the result names its city, network and ASN. MTR is the default because it repeats each hop and reports the network that owns every router; plain traceroute is one pass. “From my ISP” asks for a probe on your own network and falls back to another home or mobile connection in India when none is online.

A reverse trace runs from a probe back to your IP address, which makes your address part of a public Globalping result. Network names for each hop come from RIPE NCC RIPEstat. If you paste the output of your own tracert or traceroute, the page reads it in your browser; only if you press Look up network owners does it send the public router addresses to RIPEstat for their network names.

What Is My IP Address?

The IP address page asks Cloudflare’s speed test server for your address, network and approximate location, and asks Cloudflare’s 1.1.1.1 and ipify separately for your IPv4 and IPv6 addresses. When one of them fails, our own server, RIPEstat or ipinfo.io fill the gap. Location from an IP address is approximate and often shows your provider’s nearest hub, not your town.

The CGNAT check compares the WAN address you copy from your router with your public address, entirely in your browser; an address in 100.64.0.0/10 means carrier-grade NAT. The WebRTC check contacts the STUN servers of Cloudflare and Google only when you press its button.

IPv6 Test

The IPv6 test asks two servers that only speak IPv4 (1.1.1.1 and api.ipify.org) and two that only speak IPv6 (2606:4700:4700::1111 and api6.ipify.org) which address they see, giving each 4 seconds. A server that speaks both, api64.ipify.org, shows which protocol your browser chose; the browser decides, not the page.

Round-trip times come from a warm-up and five timed HTTPS requests to each Cloudflare address. The readiness score is ours: 10 when IPv6 works and is preferred, 7 when it works but IPv4 is preferred, 3 when there is no IPv6. The test uses well under 100 KB.

Internet Stability Test

The internet stability test is the monitor this site grew from. While its tab is open it sends a tiny request every 3 seconds to Google and Cloudflare’s 1.1.1.1, and every 30 seconds a round of small requests to a list of 27 well-known services, from Google and YouTube to WhatsApp and Amazon India. Drops, slow periods and failing services are logged with times, and you can export a report for your provider.

Samples are stored only in your browser’s IndexedDB: raw samples for 7 days, hourly summaries for a year. A sleeping computer or a browser that slows background tabs leaves gaps in the record. A request that fails is shown as a failed request, not as packet loss. Speed tests are off unless you switch them on, and each one then uses up to 37 MB.

How much internet speed do I need?

The speed calculator multiplies each activity in your home by the speed its app maker recommends, links that source on the row, adds the rows and then adds 25% headroom, which is our own choice. Rows with no official figure, such as gaming and big downloads, are marked as our estimate. The calculator makes no network requests.

Help and FAQs

The help page collects answers about speed, ping, DNS, traceroute and IP results, a glossary and an explanation of why different speed tests disagree. The good, fair and poor bands it shows are our own thresholds unless a source is named.

Your data

Results, settings and history stay in your browser, and there are no cookies, analytics or accounts. The privacy policy lists, test by test, which outside servers see your IP address and how to delete what the site stored on your device.

What a browser can and cannot measure

A web page can only use the tools the browser gives it: HTTPS requests, WebSockets, WebRTC and timers. That is enough for speed, HTTP round-trip time, DNS-over-HTTPS and your public address, but several classic network tests are out of reach.

What runs in your browser and what runs elsewhere
MeasurementIn a browser?What we do instead
Download and upload speedYesHTTPS transfers to Cloudflare, or WebSockets to M-Lab
Ping (ICMP)NoHTTPS round-trip time from your browser; real ICMP ping on a named Globalping probe
Traceroute and MTRNoRun on a Globalping probe on your network or in a city you pick
DNS on port 53 (what your phone and router use)NoDNS-over-HTTPS to public resolvers
Packet loss (UDP)Only through a relay serverShown only when measured that way; otherwise “Not measured”
Wi-Fi band, channel and signalNoYou label the room and connection type yourself
5G or 4GNoCheck the icon in your phone’s status bar
Your exact locationNo (we never ask for GPS)Approximate location from the IP address

Browsers also slow down timers in background tabs and on low battery, so keep a test’s tab in front while it runs. For the reasons tests from different sites give different numbers, see why speed tests differ.

Outside services we use

Every test talks directly from your browser to the services below. Each one sees your IP address, as any website you open does. We do not send them anything else about you, and we have no agreement with any of them beyond their public terms.

Services, the pages that use them and what they see
ServiceUsed byWhat it seesIts policy
Cloudflare speed test servers and 1.1.1.1Speed tests, bufferbloat, video call, ping, IP and IPv6 pagesYour IP address and the test trafficCloudflare privacy policy
Measurement Lab (M-Lab)M-Lab speed test and the optional second test on speed pagesYour IP address and result, published as open dataM-Lab privacy policy
Globalping (operated by Volentio JSD Limited)Traceroute, DNS trace, ICMP ping, Jio and Airtel tracesYour IP address and the target you type; measurement data is publicGlobalping terms of use
RIPE NCC RIPEstatNetwork names for hops, ISP lookup fallbackYour IP address or the network numbers looked upRIPEstat terms and conditions
ipinfo.ioLast fallback for your ISP and locationYour IP addressipinfo privacy policy
Public DNS resolvers (Cloudflare, Google, Control D, DNS.SB, CleanBrowsing, AliDNS, DNS4EU)DNS speed test and DNS lookupYour IP address and the names looked upListed in our privacy policy
ip-api.com“Which resolver am I using?”, only when you clickYour IP address and your resolver’s addressListed in our privacy policy
Amazon Web ServicesPing and gaming ping testsYour IP addressAWS privacy notice
STUN servers of Cloudflare and GoogleWebRTC check on the IP page, only when you clickYour IP addressCloudflare and Google privacy policies
ipifyIP and IPv6 pagesYour IP addressipify website

The ping test also sends small requests to Cloudflare, Google and AWS (plus any address you add), and the stability test to 27 well-known sites such as Google, YouTube and WhatsApp. These sites see your address as they do when you visit them. Hop network names use RIPEstat data, which we credit in the footer of every page.

How much data each test uses

Speed tests use real data, which matters on a mobile pack. Every speed page shows the most a test can use before you press Start and the data it really used afterwards. The measured figures below come from our own runs on 30 Sep 2026; the Data saver and Standard runs were on a line of about 200 Mbps.

Most a test can use, and what we measured
TestMost it can useMeasured in our runs
Speed test, Data saver37 MB35 MB; both directions reached the cap in about a second
Speed test, Standard217 MB210 MB in about 14 seconds
Speed test, Gigabit876 MBMeant for unlimited fibre; not measured on a mobile pack
Mobile speed test10, 25 or 50 MBThe cap you choose; the page counts the MB live
M-Lab test, Data saverabout 35 MB34.7 MB
Ping test, five serversno capabout 0.3 MB per minute
Gaming ping testno capunder 0.1 MB for 11 regions
DNS speed testno capabout 12 KB for 49 lookups
IPv6 testno capunder 100 KB
Internet stability test, default settingsno capabout 13 MB per hour (about 0.3 GB a day); each optional speed test adds up to 37 MB

A slower line uses less, because each direction also stops at a time limit. Data saver stops at its cap before the speed settles on lines faster than about 60 Mbps download (25 Mbps upload), and Standard above about 200 Mbps, so the result is a lower limit; use Gigabit on fibre if your plan has no data limit. More on this in Help: data use.

Accuracy and limits

Every result is one measurement, from one device, to one server, at one moment. Your Wi-Fi, other people using the line, the time of day and the distance to the server all change it. Run a test two or three times, and once on a cable if you can, before you draw a conclusion.

  • Different testers give different numbers. Each uses its own servers, number of connections and maths. None of them is simply right; they measure slightly different things.
  • Some results are lower limits. When a test stops at its data cap, or at its time limit while the speed is still rising, the number has “≥” in front and the page says why next to it.
  • Browser timings include some overhead. HTTP round-trip times sit a little above ICMP ping, and background tabs can be slowed by the browser.
  • Remote probes are not your device. A Globalping probe shares your network or city, not your home Wi-Fi, so its route and ping can differ from yours.
  • Location and ISP names come from databases. They are approximate and can be out of date, especially behind a VPN or CGNAT.
  • Ratings and grades are ours. Quality ratings, the A to F grade, the gaming verdict and the IPv6 score are our own thresholds, not a standard or a regulator’s figure. Where a band follows an app maker’s published figure, the page links it.

For an official record to show your provider in India, also run TRAI’s MySpeed test; our figures are a guide, not a regulatory measurement.