time-series-forecasting google-research foundation-models ai-development-tools

Google's 200M-parameter time-series foundation model with 16k context

GitHub - google-research/timesfm: TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting.

github.com

March 31, 2026

2 min read

🔥🔥🔥🔥🔥

62/100

Summary

TimesFM is a pretrained time-series foundation model developed by Google Research for time-series forecasting. The latest model version is TimesFM 2.5, with archived versions 1.0 and 2.0 available, and it can be accessed through the TimesFM Hugging Face Collection and is integrated into BigQuery as an official Google product.

Key Takeaways

TimesFM 2.5 is a pretrained time-series foundation model developed by Google Research for time-series forecasting, featuring 200 million parameters.
The model supports a context length of up to 16,000 and continuous quantile forecasting up to a 1,000 horizon.
TimesFM 2.5 includes new forecasting flags and has removed the frequency indicator from the previous version.
The inference API has been upgraded alongside the model, with plans to add support for a Flax version for faster inference.

Read original article

View story →

Coverage & discussions

Hacker News · github.com

Google's 200M-parameter time-series foundation model with 16k context ↗

Community discussions

Hacker News discussion ↗

Historical count: 5 · 5 saved in sample · Historical sample

Analysis ready

Community Sentiment

Mixed

Positives

The model's 200M parameters and 16k context length suggest it could handle complex time series data, potentially improving forecasting accuracy across diverse applications.

Concerns

The concept of a general time series model raises skepticism, as it's unclear how it can reliably predict vastly different phenomena like egg prices and global inflation.
Concerns about the model's interpretability persist, with users questioning how to trust predictions without clear explanations of the underlying reasoning.