Current conditions, an hourly strip and a 7-day forecast for your location or any city — with feels-like, wind, humidity, UV, sunrise and sunset.
—
🌤️ Data from the free, keyless Open-Meteo API. Your location is used only to fetch the forecast — never stored or uploaded.
One tap for your location, or search any city worldwide. Save the spots you check often — home, work, family — and switch instantly.
See exactly when the rain starts or the temperature drops with a clean 24-hour strip — handy before a commute, run or school run.
No sign-up, no notifications nagging you, no selling your location. Just the forecast, fast.
Modern forecasts come from numerical weather prediction: the atmosphere is divided into a three-dimensional grid, current conditions are estimated from observations, and the physics governing air movement, heat and moisture is integrated forward in time. The major global models — the ECMWF's system and the American GFS among them — run several times a day on some of the largest supercomputers in existence.
Accuracy has improved substantially and steadily. A five-day forecast today is about as reliable as a three-day forecast was in the 1990s, an improvement often described as gaining roughly a day of skill per decade. Beyond about ten days, skill degrades toward climatology — what is normal for the date — rather than genuine prediction.
The limit is fundamental rather than technical. The atmosphere is a chaotic system, so tiny differences in initial conditions grow exponentially — the observation that gave rise to the phrase "butterfly effect", from Edward Lorenz's work in the 1960s. Better models and more observations push the horizon out; they cannot remove it.
"30 percent chance of rain" is the most misunderstood figure in a forecast. It does not mean rain for 30 percent of the day, nor over 30 percent of the area. In the most common definition it is the probability of measurable precipitation at any given point in the forecast area — sometimes expressed as confidence multiplied by expected areal coverage.
So a 30 percent probability means that if this situation recurred many times, measurable rain would fall at your location in about three of ten cases. It says nothing about intensity or duration: a 90 percent chance of a brief drizzle and a 30 percent chance of a thunderstorm are different risks entirely.
Ensemble forecasting is what produces these probabilities. The model is run many times with slightly varied initial conditions, and the spread of outcomes indicates confidence — tightly clustered runs mean high confidence, widely divergent ones mean genuine uncertainty. A forecast presented as a single value hides that information.
"Feels like" temperature combines actual temperature with wind and humidity. Wind chill applies in cold conditions, modelling accelerated heat loss from exposed skin; the heat index applies in warm conditions, modelling reduced evaporative cooling when humidity is high. Both are estimates for a typical person in typical clothing, not measurements.
Forecasts are also produced for a grid cell, which may be several kilometres across, so local terrain, coastline and urban effects are not fully captured. Mountain and coastal areas are where the difference between the forecast and what happens outside is largest.
For anything where weather affects safety — flying, sailing, mountain activity, or severe weather — use your national meteorological service directly rather than a general-purpose display. Official warnings carry information that consumer forecasts summarise away, and they are issued by forecasters interpreting the models rather than by an automated feed. This page shows a general forecast and is not a source for safety-critical decisions.
Open-Meteo blends leading national weather models (the same data professional apps use). Current conditions and the next 24–48 hours are most reliable; accuracy naturally decreases further out.
Apps use different models and station blends, and "feels like" formulas vary. Small differences are normal; trends (warming, rain timing) matter more than the exact degree.
The last loaded forecast stays on screen, but fetching a fresh one needs a connection. Your saved cities and unit choice are remembered between visits.
The probability that measurable rain falls at a given point in the area during the period — not that it rains for 30 percent of the time or across 30 percent of the area. It says nothing about intensity.
Usefully to about five to seven days, degrading toward climatology beyond ten. The atmosphere is chaotic, so small errors in initial conditions grow exponentially — a fundamental limit rather than a technical one.