📶 Internet Speed Explorer
How fast is the internet, really?
Not the ISP's advertised "up to" number — the median of 13.1 million real speedtests people ran near each city center in Q1 2026. Fixed broadband (the wired/Wi-Fi line your apartment or café has) and mobile (the SIM in your phone or hotspot) are measured and rated separately, because a fiber city can have weak cellular and vice versa — and if you work from anywhere, you'll live on both. 146 of 161 measured cities clear 100 Mbps on a typical fixed connection.
Every dot is a city (log scales — distances near the origin are bigger than they look). Dashed lines mark 100 Mbps, the ✅ threshold, on both axes. Click a dot for detail. Moscow and St. Petersburg are not plotted — Ookla publishes no open data for Russia.
Both connection types are rated on median download speed: ✅ fast (≥ 100 Mbps) · ⚠️ moderate (25–99) · 🚨 slow (< 25). The first chip is fixed broadband, the second is mobile. An ≈ low sample badge means relatively few tests were run there this quarter — the median is directional, not precise.
Data last updated · Source: Ookla Open Data, Q1 2026 performance tiles (Speedtest® by Ookla, © Ookla)
How this dashboard works
Every number on this page is measured, not advertised. Ookla (the company behind Speedtest.net) publishes an open dataset of aggregated speedtest results on a global grid of ~600 m map tiles. For each of the 161 measured cities we take every tile within 10 km of the city center and compute the test-count-weighted median download, upload and latency — separately for fixed broadband (wired and Wi-Fi connections) and mobile (cellular). This quarter that's 13.1 million individual tests across the 163 cities. Medians resist being skewed by one office park with 10-gigabit fiber, and weighting by test count keeps a barely-used suburban tile from outvoting the dense center.
Two honest caveats. First, speedtest users skew toward people checking a connection — new setups, problems, upgrades — so samples aren't perfectly representative. Second, a citywide median can't see your building: a fiber city is full of old walk-ups on DSL, and a slow city usually has neighborhoods with excellent fiber. Treat these numbers as the odds you're playing, then test the actual apartment before signing a lease. Cities with an ≈ low sample badge (fewer than 300 tests this quarter) are directional only.
Where does this data come from?
Ookla's open dataset of real Speedtest results (Speedtest® by Ookla Global Fixed and Mobile Network Performance Maps, © Ookla, used under CC BY-NC-SA 4.0). We aggregate the Q1 2026 quarter: every result within 10 km of each city center, as test-count-weighted medians. Nothing is self-reported or taken from ISP marketing.
What counts as "fast enough" for remote work?
Video calls need surprisingly little download — 5–10 Mbps — but they punish high latency and weak upload. For a comfortable full-time remote setup: 50+ Mbps down, 10+ Mbps up, and ping under 50 ms. Our ✅ 100 Mbps bar is deliberately conservative: at that level you stop thinking about the internet entirely, including 4K calls, big uploads and cloud backups.
Why show fixed and mobile separately?
Because they fail independently and you depend on both: fixed is your apartment and café Wi-Fi; mobile is your hotspot backup, your maps, your on-the-road connection. Some cities pair world-class fiber with mediocre cellular (much of Southeast Asia's fiber hubs); others are mobile-first, where 5G beats the aging copper (Tbilisi is a striking example). The scatter chart puts every city on both axes at once.
Why is the median so different from speeds ISPs advertise?
Advertised speeds are best-case plan maximums; the median is what people actually got, including congested evenings, old routers, and Wi-Fi three rooms from the modem. The gap between the two is itself useful information about a market.
Why is there no data for Moscow and St. Petersburg?
Ookla ceased operations in Russia in 2022 and publishes no open-data tiles for the country. We keep both cities listed with a "no published data" flag rather than pretending they don't exist — but we have nothing measured to show, and the gap says nothing about actual speeds on the ground there.
What does the "low sample" badge mean?
Fewer than 300 speedtests (or 10 map tiles) for a connection type this quarter — common in smaller destinations. The median is still the best available measurement, but treat it as directional and test locally before relying on it.