Contact for pricing · Free trial🇺🇸🇩🇪
Process optimization for steel, chemicals and cement with explainable machine learning and setpoint advice.
TimesFM 2.5 overview and key facts (October 2026)
Open foundation model from Google Research for time series forecasting with quantile output, also available in BigQuery.
Free
What is TimesFM 2.5?
TimesFM is a pretrained time series foundation model from Google Research. Analysts use it to forecast business metrics such as sales, traffic or demand, either in their own Python environment or through BigQuery.
Version 2.5 has 200 million parameters in a decoder-only architecture, takes up to 1,024 time steps as context and forecasts up to 256 steps ahead. The model card recommends reinstalling the latest package after an October 2025 update. Hugging Face counted about 1.6 million downloads in the last 30 days as of September 28, 2026.
TimesFM 2.5 AI features and use cases
- Forecasts business metricsTimesFM predicts future values of a time series from its history without training a model per series. Analysts get forecasts for many products or locations with little setup.
- Shows ranges, not just a numberNext to the mean forecast, the model returns quantiles from the 10th to the 90th percentile. Planners see how uncertain a forecast is and can plan buffers.
- Forecasts inside the data warehouseGoogle supports TimesFM in BigQuery through AI.FORECAST. Data teams forecast where their data already lives, without moving it to a separate system.
What sets TimesFM 2.5 apart
- Officially supported in BigQuery through AI.FORECAST, next to the open weights.
- Released under Apache-2.0 with 200 million parameters.
Where TimesFM 2.5 is strongest
United States Available internationally
TimesFM is an open model from Google Research under Apache-2.0 and can be used worldwide without country ties.
Questions to ask before you buy TimesFM 2.5
The context length of 1,024 steps limits how much history the model sees, so check that it fits your data frequency. Compare forecast quality with your current method first. If you use BigQuery, clarify query costs and the data location of your project.
TimesFM 2.5 reviews
No reviews yet. Be the first to share how your team uses it.
Every review shows its verification level. We read each one before it goes live. How we check reviews
Questions about TimesFM 2.5
Is TimesFM free?
The model weights are free under Apache-2.0. You pay only for compute, or for BigQuery usage if you use AI.FORECAST.
How far ahead can TimesFM 2.5 forecast?
According to the model card, it forecasts up to 256 steps from a context of up to 1,024 steps.
Does TimesFM provide uncertainty ranges?
Yes. It returns the mean forecast and quantiles from the 10th to the 90th percentile.
TimesFM 2.5 update historyno updates yet
No changes recorded for TimesFM 2.5 yet.
Changes the vendor made and confirmed user reports, each checked against the vendor's own pages. We review every tool monthly.
Alternatives to TimesFM 2.5
Tools in the same category (Forecasting) with the most overlapping markets.
Free🇺🇸
Open foundation model from Amazon for zero-shot time series forecasting, supporting multiple series and covariates.
https://jolene.pro
Vendor verifiedFreemium · Free trial🇺🇸🇨🇦🇬🇧and 4 more countries
Data analysts use this tool to process multi-sheet Excel and CSV files, generate charts, and trace results back to their source cells.
Contact for pricing🇧🇪
Smart building platform for workplace, energy and maintenance with AI anomaly detection, hosted in the EU.