DEV Community

Cover image for Battery Forecast — Local AI Battery Predictions for Windows
Mattboxx
Mattboxx

Posted on Originally published at mattboxx.github.io

Battery Forecast — Local AI Battery Predictions for Windows

I created Battery Forecast, an open-source Windows adaptation of Matteo Bonfanti’s Till Dusk, powered by Mistral AI’s Timestral-28M model.

Instead of relying only on Windows’ battery estimates, it uses your local battery history to predict:

  • How long your battery will last based on your usage.
  • When charging will reach your target (e.g. 80%).
  • Your expected battery level later in the day.
  • Battery trends over the next 24 hours.

Features:

  • Fully offline AI inference — no cloud or account required.
  • Battery history and health monitoring.
  • System tray app with customizable alerts.
  • Estimated time ranges and interactive charts.
  • Supports Windows 10/11 (x64 and ARM64).

The model runs locally on your CPU and uses historical data without retraining.

Note: Forecast accuracy is still being evaluated, and charging predictions remain experimental.

GitHub: https://github.com/Mattboxx/battery-forecast

Website: https://mattboxx.github.io/battery-forecast/

I’d love to hear your feedback, especially from people willing to test the predictions on their own laptops!

Top comments (0)