Internet connection test
See how your connection responds when it is quiet and when a transfer is running.
Keep this page visible until the test finishes.
Ready when you are
Measuring your connection
Baseline → Download → Upload
No test has run yet.
Your result
Local onlyYour completed reading
Everyday activities Browsing, gaming, calls and streaming
Change one condition and repeat. This page keeps your last two completed readings for comparison until you reload or leave.
Detailed results — show more data Probe chart, jitter, failures, transfers & device context
Measurement method & evidence limits
Baseline means before this test's transfers; other traffic can still be present. HTTP timing includes this device, the network and endpoint. Failed requests are excluded from percentiles and counted separately; they are not packet loss. Jitter is the mean change between adjacent successful probes.
Your JSON keeps the raw measurements, context, interpretation and publication state. Pending or unknown publication may already be public. It excludes publication credentials and exact coordinates.
Latency first. Speed included. How we measure ↗
How long you wait.
Time for an HTTP reply, in milliseconds. Lower is better. We compare before transfers, during downloads and during uploads.
How much you move.
Download and upload throughput, in Mbps. Higher moves files faster. A fast connection can still have slow replies when it gets busy.
Numbers, explained.
See whether replies slowed under load, what could affect everyday use, and how much evidence the test collected.
Read the latency guide ↗Voluntary public observations
Compare connections
World observations
Filters and method
Every filter is exact. Browser execution describes this device, but does not establish local or cloud placement. Historical unknown placement and publication-session attribution remain separate from measurement facts.
Each displayed value is a server-calculated median of eligible per-test values. Missing values stay unavailable. Five eligible tests are required per metric; this editorial floor does not establish confidence or anonymity. HTTP failures are not packet loss. Throughput is achieved payload rate, not capacity or retail service availability.
Everyday method and provenance
| Area and observed network | During download HTTP p95 | Evidence |
|---|