Download speed
0.00Mbps
Peak — · Avg — · 0 MB
The speed machine
Run a full diagnostic
Three steps: choose a server, start the test, read your score.
- 1Pick a server, or leave it on auto
- 2Press start and keep the tab open
- 3Read your score, share or export
Idle
0.00
Mbps
Ready to test
- Initialise
- Network
- Ping
- Download
- Upload
- Analyse
- Report
Google and Microsoft endpoints do not send CORS headers, so only latency can be measured against them — throughput automatically falls back to the nearest CORS-enabled edge.
Download uses GET {base}/__down?bytes=N, upload uses POST {base}/__up.
Measured live in your browser: idle latency uses zero-byte probes with server processing time (Server-Timing) subtracted; download uses up to six parallel streams across 1–50 MB payloads and upload three streams across 1–10 MB payloads; throughput is computed only from the steady-state window after ramp-up, with statistical outliers discarded.
Upload speed
0.00Mbps
Peak — · Avg — · 0 MB
Ping (latency)
0ms
Min — · Max — · Median —
Jitter
0ms
Std dev — · Probe loss —
Network quality score
—
/ 100
Awaiting test
Score blends download, upload, latency, jitter and packet stability.
Connection & device details
IPv4 address: Detecting…
IPv6 address: Detecting…
ISP / organisation: Detecting…
Location: Detecting…
Test server: Auto-selecting…
Streams: 6 down / 3 up
Connection type: Checking…
Effective type: Checking…
Browser RTT estimate: Checking…
Browser downlink estimate: Checking…
Data Saver: Checking…
Browser: —
Operating system: —
Screen resolution: —
Data transferred: 0 MB
Time taken: —
Measurement confidence: —
Last test: —
Export & share
Reports are generated entirely inside your browser — nothing is uploaded. The PDF option opens your browser’s print dialog; choose “Save as PDF”.
Real-time analytics
Live telemetry
Every sample is plotted as it happens, so you can spot ramp-up behaviour, throughput dips and latency spikes instead of one flat average.
Throughput timeline (Mbps)
Latency timeline (ms)
Your last 10 tests
Charts are unavailable because the Chart.js CDN could not be loaded. All numeric results above remain fully accurate.
Historical reports
Test history
Your last 20 tests are kept in this browser only — never on our servers.
| Date | Download | Upload | Ping | Jitter | Score | Quality | Server | Actions |
|---|
No saved tests yet — run your first speed test to start the collection.
Core features
Serious accuracy, simple interface
A measurement pipeline built for fibre, cable, DSL, 5G and satellite links — and honest reporting whenever a browser API simply cannot provide data.
Ultra-accurate testing
Parallel streams saturate modern links, slow-start is discarded, and speed comes from differential byte accounting inside a steady-state window.
Real-time analytics
Live Chart.js telemetry plots every throughput and latency sample as it is captured, exposing fluctuation an average would hide.
Privacy first
Only throwaway random bytes cross the wire. Results and history live in your browser’s localStorage and never reach us.
No installation
Nothing to download, no extension, no Flash or Java. Open the page and press start — that is the entire workflow.
Global testing
Measurement runs against a global anycast edge, so you hit the nearest point of presence automatically — or pick a server yourself.
Cross platform
Identical engine on Chrome, Firefox, Edge, Safari and Opera, with XHR fallbacks where Streams or AbortController are missing.
Network diagnostics
Ping distribution, jitter, probe loss and packet stability are reported separately so you can localise the actual fault.
Bandwidth analysis
Peak, average and steady-state throughput plus a confidence figure derived from sample variance — not one lucky burst.
Historical reports
Twenty stored runs, a comparison chart and JSON/CSV export make it easy to prove degradation to your provider.
Enterprise APIs
Embeddable widgets and a measurement API for ISPs, MSPs and NOC dashboards — talk to our engineering team.
Understand your results
What each metric really means
Throughput only matters in context. Here is how engineers read the very same numbers you just measured.
Ping
Time for one packet to reach the server and come back. Under 30 ms feels instant; above 100 ms adds noticeable delay to calls and games.
Latency
The broader delay budget, including queuing inside your router. Idle latency is measured with no load; loaded latency rises when the link is saturated.
Jitter
Mean variation between consecutive round trips. Above roughly 30 ms, voice and video degrade even when average speed looks healthy.
Bandwidth
The theoretical capacity your plan sells you — the maximum the pipe can possibly carry.
Throughput
What you actually achieve after protocol overhead, Wi-Fi loss and congestion. This is what the gauge shows.
Bufferbloat
Excess buffering that inflates latency during heavy transfers. If ping under load is far worse than idle ping, enable smart queue management (SQM/fq_codel).
Packet loss
Packets that never arrive, forcing retransmission. Browsers cannot count raw loss, so we report probe loss — failed HTTP probes — as a proxy.
Connection quality
Our 0–100 score weights download 40%, upload 20%, ping 25%, jitter 10% and stability 5%.
Gaming performance
Needs low ping and low jitter far more than raw speed: 25 Mbps with 15 ms ping beats 500 Mbps with 90 ms ping.
Streaming quality
Roughly 5 Mbps per HD stream and 25 Mbps per 4K stream, multiplied by simultaneous viewers.
Video conferencing
Group HD calls want 4 Mbps up and down per participant, jitter under 20 ms and stable latency.
Benchmarks
Under 25 Mbps: browsing and SD video. 50–100 Mbps: HD streaming, multi-device work. 200 Mbps+: 4K, gaming and big downloads.
FAQ
Speed test questions
How accurate is a browser-based speed test?
Very accurate when it is engineered correctly. Circuitbrowsly opens up to six parallel download streams (a single TCP stream cannot fill a modern fibre or 5G link), discards the TCP slow-start ramp-up, computes speed from differential byte counters using the high-resolution performance clock, and discards statistical outliers. The limiting factors are then your Wi-Fi link and device CPU — exactly the path your real traffic takes.
Why is my speed lower than the plan I pay for?
Wi-Fi interference, an older router, VPNs, background downloads, shared bandwidth at peak hours and device CPU limits all reduce achievable throughput. Test again over Ethernet with other devices idle, then compare the two figures before contacting your ISP.
What is a good ping and jitter value?
Ping below 30 ms is excellent, 30–60 ms is good, 60–100 ms is acceptable, and above 150 ms hurts real-time applications. Jitter should stay under 10 ms; above 30 ms you will hear artefacts in calls.
How much data does one test consume?
Typically 60–400 MB depending on how fast your line is, because faster links transfer more bytes inside the same measurement window. The live “Data transferred” counter shows the exact amount, so mobile users can decide before starting.
Do you store my results or IP address?
No. History is written to your browser’s localStorage and can be cleared with one button. Your IP is fetched only to display it to you, and only random throwaway bytes are transferred during measurement. See our Privacy Policy for the full detail.
Why do some fields say “Unavailable in your browser”?
Because the underlying API genuinely does not exist there. The Network Information API (effective type, RTT estimate, downlink, Data Saver) is unsupported in Safari and Firefox, and some networks block IP-lookup services. We prefer stating that honestly over inventing a value.
Enterprise
Need custom speed monitoring?
Talk to our engineering team about dedicated measurement APIs, white-label widgets, and continuous multi-region monitoring for your subscriber base.
- Email: [email protected]
- Phone: +1 (800) 555-0199
- Hours: Mon – Fri, 08:00 – 19:00 (PT)
- Office: 548 Market St, Suite 89201, San Francisco, CA 94104, USA