← Discover
Discover · GitHub · Timesfm↑ +606 today

AI forecasting for time series

The Time Series Foundation Model, developed by Google Research, offers a pre-trained structure for time series forecasting. The model is designed to provide general predictive capabilities on different data sets.

Updates

  • September 8, 2026: Stars 30,360 → 31,902, latest release v3.0.0 (August 28, 2026).
  • September 3, 2026: Stars 28,335 → 30,360, latest release v3.0.0 (August 28, 2026).
  • August 31, 2026: Stars 27,185 → 28,335, latest release v3.0.0 (August 28, 2026).
  • August 2, 2026: Stars 22,167 → 27,185, latest release v2.0.2 (July 2, 2026).

What you get

  • Fast prediction with pre-trained base model
  • 16k context length support
  • Adaptation to different data sets with flexible structure

Installation

Installation via PyPI
pip install timesfm[torch]
# Or with Flax
pip install timesfm[flax]
# And when XReg is needed
pip install timesfm[xreg]
local installation
git clone https://github.com/google-research/timesfm.git
    cd timesfm

If you don't write code

🤖 Paste this into your AI agent (Claude Code · Codex · Antigravity)

I want to do time series forecasting using the TimesFM library. How can I configure the 200M parameter structure with version 2.5 of this basic model developed by Google? How should I determine max_context and max_horizon values, especially during the compile phase of the model, and in what format should I provide data input to the forecast function? Can you explain with an example code structure?

Related dictionary terms

Who it is forIt is designed for data scientists and researchers who want to make predictions on time series data.
LicenseApache-2.0

Links

TreScout did not build this tool · we found it in GitHub trends and wrote it up. This page describes the repository as of 2026-06-18: The star count and our text belong to that day, the repository may have changed since. Check the repository link for the current state. This page was machine-translated from the Turkish original · the Turkish version prevails.