Measure ping, time to first byte and download speed through any proxy, with a clear account of what the numbers do and do not prove.
One 1 MB download is pulled from this server through your proxy. That gives you throughput on this route, at this moment, from a server in Germany — not the proxy's ceiling. A proxy that is fast to us can be slow to the site you actually care about, so treat this as a sanity check rather than a benchmark.
Speed tests move real data, so they are limited to a few every five minutes per visitor.
Connect time is how long the handshake took — effectively the round trip between this server and the proxy. It is the closest thing here to a ping, and it is dominated by physical distance.
Time to first byte is how long the proxy took to start answering after the request went out. A proxy with a fast connect time but a slow first byte is usually overloaded rather than far away.
Download speed is throughput over the body of the transfer, in megabits per second. The payload is random data, not zeros, because compressible test data makes every proxy look identically quick.
Providers quote the best case: a short hop, an idle gateway, a large file. This measures one route — from a server in Germany, through your proxy, at this moment — so it will usually read lower, and that is the point. What it is good for is comparing two proxies under identical conditions on the same afternoon.
Residential and mobile proxies will look slow next to datacentre ones. That is inherent: the traffic is crossing somebody's home or mobile connection. A residential proxy at a few megabits is performing normally.
Each test moves real data through this server, so it is limited to a few runs every five minutes per visitor. If you need to measure a large pool, do it from your own machine — the results will be more useful anyway, because they will reflect your route rather than ours.