Businesses can now forecast daily sales with more of the context that actually moves demand, because Google TimesFM-3 reads related products, weather, past foot traffic, and planned discounts alongside the sales history itself. Google Research released the model on September 12, 2026, and it works zero-shot, so it produces forecasts on a new dataset without extra training. It is the first version in the family to handle several data series at once, which is what makes the added context possible.
What does Google TimesFM-3 do?
TimesFM-3 forecasts time series such as daily sales figures. Real forecasts rarely depend on a single variable, and Google illustrates the point with a retail chain predicting ice cream sales. A useful forecast there should factor in related products like waffle cones or syrup, along with past foot traffic, weather, discount campaigns, and holidays.
The model has 330 million parameters and was trained on real and synthetic time series totaling more than one trillion data points, according to Google. It handles three types of supplementary data. It predicts multiple related variables at once, such as different ice cream flavors. It incorporates factors known only for the past, such as historical foot traffic. It also uses known future events like planned discounts or weather forecasts. Rather than a single point estimate, TimesFM-3 outputs nine values per time step to capture the range and uncertainty of each prediction.
How does the model read related data?
TimesFM-3 is built on a Transformer, the same base architecture as its predecessors. It groups 32 consecutive data points into a single patch and normalizes each series to a common scale, so measurements of very different magnitudes can be compared directly.
The model processes data in two alternating directions. Along the time axis, it looks for patterns within a single series and draws only on past values to avoid leaking future information. Across series, it compares all variables at a given point in time and learns how they relate, which lets it pick up on effects like how a discount on one product changes sales of another.
One-pass forecasting replaces step-by-step prediction
Earlier versions predicted the future one block at a time, an approach Google describes as slow and compute-heavy, and one where errors compounded as each prediction built on the last. TimesFM-3 marks all future time steps as blanks and fills them in a single pass.
The ice cream example shows the payoff. A model that only knows past sales continues the usual weekly pattern and stays blind to planned promotions. Once TimesFM-3 receives the discount schedule, it learns from history how much promotions boost demand and expects roughly 20 percent more units on each promotion day.
How does TimesFM-3 perform on benchmarks?
On Gift-Eval, FEV-Bench, and Time, TimesFM-3 ranks first among all pretrained forecasting models in both point accuracy and uncertainty calibration, according to Google. Its competitors include Amazon’s Chronos-2, the Toto-2.0 family, and Google’s own TimesFM-2.5.
Even limited to a single variable, TimesFM-3 matches or beats the field, and adding more data widens the gap. On Gift-Eval it leads by a wide margin. On FEV-Bench, Chronos-2 comes close to TimesFM-3’s single-variable mode but falls well behind the full version. On the Time benchmark, the Toto-2.0 family follows TimesFM-3, while TimesFM-2.5 lands well back.
Where can you use TimesFM-3?
TimesFM-3 is available on GitHub and Hugging Face, and Google plans to add it to BigQuery in the coming weeks. TimesFM-2.5 currently handles single-variable forecasting in BigQuery through the AI.FORECAST command. Since the family launched in 2024, Google says it has been deployed in retail, finance, manufacturing, healthcare, and the sciences.
All versions through TimesFM-2.5, released in September 2025, could process only one data series at a time, so the multivariate support in TimesFM-3 is a significant step for the family. Google is also building forecasting models beyond time series. In early August, Google DeepMind released WeatherNext Cyclones, an open-source system for tropical cyclones that predicts storm tracks and intensity about a day further out than leading operational models.
FAQ
What is Google TimesFM-3?
TimesFM-3 is a forecasting model from Google Research that predicts time series such as daily sales. It has 330 million parameters, was trained on more than one trillion real and synthetic data points, and works zero-shot, meaning it needs no extra training for a new task.
How is TimesFM-3 different from earlier versions?
All versions through TimesFM-2.5 could process only one data series at a time. TimesFM-3 handles multiple related variables at once, factors known only for the past such as historical foot traffic, and known future events such as planned discounts and weather forecasts. It also fills all future time steps in a single pass instead of predicting one block at a time.
Where can I access TimesFM-3?
TimesFM-3 is available on GitHub and Hugging Face, and Google plans to add it to BigQuery in the coming weeks. In BigQuery, TimesFM-2.5 currently handles single-variable forecasting through the AI.FORECAST command.
Related coverage
This article summarizes reporting from the-decoder.com. See our editorial disclaimer for how our articles are produced.
Run a free scan to see your AI Visibility Score, SEO rating, and local citation accuracy.