VPN speed test
Measure your line, turn the VPN on, measure again. The answer is not a number of megabits — it is how much of your own connection survived, and whether that is normal.
Free, no sign-up, nothing stored. Part of our VPN tool set · data from tests run September 4, 2026.
Connect to a nearby server for the second run. Testing against a server in another hemisphere measures the distance, not the VPN, and it is the single most common reason people conclude their provider is slow.
- Measure with the VPN off — that is your real line
- Connect to a nearby server and measure again
- Compare: the figure that matters is what survived
- 4 parallel streams
- Our own test method
- Compare against 940 Mbps
- Download, Mbps
- —
- Upload, Mbps
- —
- Latency and jitter
- —
- Speed retained vs unprotected
- —
Same method as our own rig: 3 runs per server, median reported, first quarter of each run discarded.
- 1Measure with the VPN off
- 2Connect the VPN, measure again
Start with your VPN off
A VPN speed on its own tells you almost nothing — 300 Mbps is excellent on a 400 Mbps line and poor on a gigabit one. The number worth having is how much of your line survives the tunnel, so we need the line first. Takes about twenty seconds.
About this tool
Nearly every VPN speed test is a generic speed test with a VPN logo on it: it hands you a number in megabits and leaves you to decide whether that number is good. It is not decidable. 300 Mbps is excellent on a 400 Mbps line and disappointing on a gigabit one, and the tunnel is only one of several things standing between the two figures.
Retention is the measurement
So this tool asks for two runs and reports the share of your own line that survived. That figure is comparable — between people, between providers, and against our own testing. Across the ten VPNs we measure, retention on a nearby server ran from 47% to 76% of a 940 Mbps line, and added latency from 4 to 14 ms. When your result comes back, it is placed against that range rather than left hanging.
What we do to avoid under-reporting you
Two things, because browser speed tests are easy to get quietly wrong. The transfer uses four parallel streams, since a single connection cannot fill a fast line. And the first quarter of the transfer is discarded — that is TCP slow start, when the connection is still ramping up, and averaging it in is the usual reason a browser test reports a gigabit line as 600 Mbps. Latency is a median of ten pings with stalls dropped, and jitter is the mean gap between consecutive ones, which is what you actually hear on a call.
Before you blame the provider
A result well below the range above is far more often a setting than a subscription. Connect to a server near you rather than the one still selected from last month. Set the protocol to WireGuard explicitly rather than Automatic, which quietly falls back to OpenVPN on some networks. Then run this two or three times — line speed swings enough in the evening that one pair of runs can be off by a third, which is why our own published figures are medians of three runs per server across four regions, re-measured monthly. The method is on the how we test page; the fastest VPN ranking has all ten by retention on the same line, the server location guide has the distance numbers and the protocols guide the WireGuard ones. We earn a commission if you buy through a link here, and it plays no part in any of the figures (disclosure).
Questions people ask
How much speed should a VPN cost me?
Between a quarter and a half of your line, on a nearby server. Across the ten VPNs we test, retention ranged from 47% to 76% of a 940 Mbps connection, and added latency ranged from 4 to 14 ms. A result far below that range is usually the server you picked rather than the provider you pay.
Why do I have to test twice?
Because a VPN speed on its own means nothing. 300 Mbps is excellent on a 400 Mbps line and poor on a gigabit one, so the only comparable figure is the share of your own connection that survives the tunnel. That needs a baseline, which is what the first run is for.
Why is my VPN speed test slower than a normal speed test?
Partly the tunnel, and partly the server. Encryption costs a little; distance costs a lot. The most common cause of a bad result is a server on the other side of the world that somebody selected weeks ago for one evening's viewing and never changed back. Connect somewhere near you and run it again before drawing conclusions.
Does this test upload as well as download?
Yes, and latency and jitter alongside them. Upload matters more than most speed tests imply — it is what a video call, a cloud backup and torrent seeding all depend on, and it is where some VPNs fall away much harder than their download figures suggest.
Is one test enough to judge a VPN?
No, and this is where most VPN speed comparisons go wrong. Line speed varies minute to minute, especially in the evening, so a single pair of runs can easily be off by a third. Run it two or three times at different moments. Our own published figures are the median of three runs per server across four regions, re-measured monthly.
Do you store my results?
No. The measurement runs entirely in your browser against Cloudflare's public measurement endpoint, and your baseline is kept in your own browser's storage for six hours so you do not have to re-measure your line between runs. Nothing reaches us and nothing is tied to you.
More free tools
- AI VPN finder (chat)Describe what you need in your own words and Emilia picks one, using our own test data. Ask follow-ups until you're sure.
- What is my IP address?See the IP, location and ISP every site you open can read right now — and whether a VPN is actually hiding them.
- DNS leak testYour VPN can be connected while your DNS lookups still go to your ISP. This asks our own nameserver which resolvers came looking.
- WebRTC leak testYour browser can hand out your real IP behind a working VPN. This runs the check in your browser and tells you if yours does.
- VPN finderFive questions about how you'll actually use it, scored against our test data. One recommendation, one runner-up, and why.
- VPN price calculatorWhat a VPN really costs once the intro deal ends. Pick a term, see the total and the renewal for every VPN we rate.
- VPN server location pickerWhich country to connect to for what you're actually doing — and why the answer is the nearest server more often than you'd think.