Time Series Forecasting Using Foundation Models

by Marco Peixeiro

Make accurate time series predictions with powerful pretrained foundation models!

You don’t need to spend weeks—or even months—coding and training your own models for time series forecasting. Time Series Forecasting Using Foundation Models shows you how to make accurate predictions using flexible pretrained models.

In Time Series Forecasting Using Foundation Models you will

• The inner workings of large time models
• Zero-shot forecasting on custom datasets
• Fine-tuning foundation forecasting models
• Evaluating large time models

Time Series Forecasting Using Foundation Models teaches you how to do efficient forecasting using powerful time series models that have already been pretrained on billions of data points. You’ll appreciate the hands-on examples that show you what you can accomplish with these amazing models. Along the way, you’ll learn how time series foundation models work, how to fine-tune them, and how to use them with your own data.

About the technology

Time-series forecasting is the art of analyzing historical, time-stamped data to predict future outcomes. Foundational time series models like TimeGPT and Chronos, pre-trained on billions of data points, can now effectively augment or replace painstakingly-built custom time-series models.

About the book

Time Series Forecasting Using Foundation Models explores the architecture of large time models and shows you how to use them to generate fast, accurate predictions. You’ll learn to fine-tune time models on your own data, execute zero-shot probabilistic forecasting, point forecasting, and more. You’ll even find out how to reprogram an LLM into a time series forecaster—all following examples that will run on an ordinary laptop.

What's inside

• How large time models work
• Zero-shot forecasting on custom datasets
• Fine-tuning and evaluating foundation models

About the reader

For data scientists and machine learning engineers familiar with the basics of time series forecasting theory. Examples in Python.

About the author

Marco Peixeiro builds cutting-edge open-source forecasting Python libraries at Nixtla. He is the author of Time Series Forecasting in Python.

Table of Contents

Part 1
1 Understanding foundation models
2 Building a foundation model
Part 2
3 Forecasting with TimeGPT
4 Zero-shot probabilistic forecasting with Lag-Llama
5 Learning the language of time with Chronos
6 A universal forecasting transformer
7 Deterministic forecasting with TimesFM
Part 3
8 Forecasting as a language task
9 Reprogramming an LLM for forecasting
Part 4
10 Capstone Forecasting daily visits to a blog

Get a free eBook (PDF or ePub) from Manning as well as access to the online liveBook format (and its AI assistant that will answer your questions in any language) when you purchase the print book.

460 pages, Kindle Edition

Published December 16, 2025

Reviews

[Sumit Bhattacharyya · Rating 5 out of 5]

Time Series Forecasting Using Foundation Models is a timely and practical book that focuses on how modern foundation models—primarily transformer-based architectures—are being applied to time-series forecasting in real-world settings.

The author does a solid job explaining why naïve or “vanilla” transformers historically performed poorly on forecasting benchmarks, and then methodically walks through the architectural adaptations that make large-scale models like TimeGPT, Chronos, Moirai, and TimesFM viable in practice. Concepts such as patching, positional encoding for time series, zero-shot forecasting, fine-tuning, and handling exogenous variables are explained clearly, with enough depth to be useful without becoming overly theoretical.

A clear strength of the book is its practitioner lens. The examples reflect real operational constraints—heterogeneous datasets, multiple frequencies, long horizons, and the need for generalization across domains. The discussion around when foundation models do not outperform simpler statistical or linear baselines is also refreshingly honest.

That said, readers should be aware of the author’s industry context: the book is written by a practitioner from Nixtla, the creators of TimeGPT. While this does introduce a natural emphasis on transformer-based foundation models, the technical content remains rigorous and the limitations are acknowledged rather than glossed over.

This is not a beginner’s book on forecasting, nor a replacement for classical time-series texts. It is best suited for practitioners who already understand forecasting fundamentals and want to evaluate modern foundation-model approaches critically and pragmatically.

Recommended for: applied ML practitioners, data scientists, and architects interested in modern time-series forecasting at scale.

[Avinash · Rating 5 out of 5]

This book is a timely and practical introduction to the emerging use of foundation models for time-series forecasting. Instead of focusing on building classical models from scratch, it shows how to leverage pre-trained models such as TimeGPT, Chronos, Lag-LLAMA, and others for zero-shot forecasting and fine-tuning. The strength of the book lies in its hands-on, code-oriented approach, making it especially useful for practitioners who want to experiment quickly with modern forecasting tools.

However, it is not a beginner’s book. Readers are expected to already understand core time-series concepts and be comfortable with Python. The theoretical depth is intentionally limited, and the book should be seen as a complement to traditional forecasting texts rather than a replacement. Given how new the field is, some claims around generalization and performance should be taken with a practical, experimental mindset.

Overall, this is a solid, forward-looking resource for data scientists and ML engineers who want to explore state-of-the-art forecasting workflows using foundation models, with a clear focus on real-world applicability rather than theory.

[Josua Naiborhu · Rating 5 out of 5]

clear and practical book to understand time series forecasting using foundation models. I love each foundation models mentioned for benchmarking my future time-series project.