You check your weather app every morning. It tells you it’s going to be sunny, so you skip the umbrella. Then it rains. You shrug it off — weather is unpredictable, right? Wrong. The app isn’t unpredictable. It’s just not telling you the truth.
I spent weeks building a scoreboard that tracks which weather models actually deliver. The results are uncomfortable. The best-performing model in the world — an AI-driven system called AIFS — is almost entirely absent from the apps you use every day. Meanwhile, the models that power your phone’s forecast are chosen by cost and licensing, not by accuracy. And the apps themselves? They’ve never shown you their work.
Here’s what I found: AIFS crushes the competition on temperature, wind, and rain predictions. Yet virtually no commercial weather app gives you access to it. Some might blend it in quietly, but they won’t tell you. The second-best performer? Foreca, a relatively obscure provider. The popular ones? GFS and ICON — they’re fine, but they’re not the best. And ICON is wildly regional: it’s great around the Mediterranean, terrible everywhere else.
This isn’t about atmospheric science anymore. The bottleneck to better weather forecasts isn’t atmospheric science — it’s commercial distribution. Weather apps all market accuracy, but they hide the verification that would prove it. The best-scoring model is the one users are least likely to get. That’s not a technical problem. That’s a market failure.
So I built an open-source scoreboard that finally shows you the truth. It takes forecasts from ECMWF, GFS, ICON, AIFS, Apple Weather, Foreca, OpenWeather, Visual Crossing, and others — then waits for the weather to happen and scores each one. You can see, for any city, which model is actually right. No more blind trust. Your weather app has never shown its work. Now you can check.
The data is clear: the model you’re getting isn’t the best one. It’s the one that costs the least to license. The apps that promise precision are selling you convenience, not accuracy. And the entire industry relies on you never asking for proof.
But now you have proof. The scoreboard is live, it’s open-source, and it runs in your browser. Go check your city. You might be surprised — or angry. Either way, you’ll never look at your weather app the same way again.
FAQ
Q: How can I trust your scoreboard isn't biased?
A: Because it's open-source and reproducible. Every model's forecast is pulled from public archives or direct APIs, and the scoring method is transparent. You can run the same code yourself or verify the data against the source.
Q: So which weather app should I use?
A: That depends on your location. For most regions, AIFS is the best raw model, but few apps offer it directly. If you want the best commercial blend, Foreca scores well. The real takeaway is to check the scoreboard for your city and pick an app that uses the top model — or just use the raw model data directly.
Q: Isn't weather inherently unpredictable? Why blame the apps?
A: Yes, weather is chaotic — but that's a separate issue. The point is that among the available models, some are measurably better than others. Apps could use the best ones, but they don't. They choose cheaper or more restrictive licenses, then market themselves as 'accurate.' That's a choice, not a limitation of science.