Speedtest CLI vs browser

[kss]

I am running OPNsense 25.1.10 and noticed that the Speedtest CLI is reporting much lower speeds than the browser version to the same server (NovoServe LLC
Edison, NJ), few seconds apart.

Results

Browser - Download 511.41 Mbbs; Upload 569.03 Mbps; Ping 11 ms; Download Latency 17 ms; Upload Latency 45 ms
CLI -     Download  21.55 Mbbs; Upload   9.17 Mbps; Ping 27 ms; Download Latency ?;     Upload Latency ?

Empirically, the network performance seems fine, even when running multiple streams.

Is it just me? Any idea what might be going on or where I should start looking?

(I incorrectly posted this earlier to 24.x forum, sorry)

[meyergru]

Theorectically, the difference could be that your browser runs on a client, but I gues you are using the CLI version on OpnSense itself? You can well have problems that allow your clients to have different routes or settings than OpnSense.

For example, when you only use IPv6 prefixes (IA_PD), but no IA_NA and your OpnSense has no IPv6 of its own, it could happen that it can only use IPv4, but no IPv6. Depending on the speed of IPv4 vs. IPv6 on the way to the test server, the results can differ. Same goes for policy-based routes.

IDK what type of traffic Speedtest uses, though. A packet trace would show any differences. Speedtest is not exactly transparent about what IP the server is.

[kss]

Yes, I am using the CLI version on OpnSense itself. I do expect some difference in speed given the number of variables, but not 25x worse numbers for the CLI.

The speedtests were performed against the same server IP address in Edison, NJ in quick succession.

I am not using IPV6, although the WAN does have an IPV6 address assigned to it.