Google Research's pretrained time-series foundation model for zero-shot multivariate forecasting with covariates.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
TimesFM 3.0 is the latest version of Google Research's TimesFM (Time Series Foundation Model) family for time-series forecasting. Released in August 2026, it adds native multivariate forecasting and flexible covariate support (past-only and past-and-future), alongside univariate forecasting, without per-task tuning. Creator coverage describes it as a roughly 330M-parameter model pre-trained on over a trillion time points.\n\nThe README reports rank #1 overall on fev-bench (100 tasks) and the TIME benchmark (50 datasets, 98 tasks), and rank #1 among foundation models on GIFT-Eval. These are vendor-reported results. Earlier versions (1.0 to 2.5) remain available, with 2.5 using 200M parameters and up to 16k context.\n\nThe code is Apache-2.0, but TimesFM 3.0 weights are under a separate non-commercial license that restricts them to non-commercial, non-production use. Commercial and production use is permitted through authorized Google Cloud services such as BigQuery ML, which is also integrated with Google Sheets, and via Vertex Model Garden. The open version is not an officially supported Google product.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Open-source, zero-shot, multivariate time series foundation model with 330M parameters, pre-trained on over a trillion time points. Ranked first among time series foundation models on several benchmarks.
Closest neighbours by meaning, not just by category.
Model · Liquid AI
Liquid AI's LFM3 family of small multimodal foundation models for on-device, edge and cloud use.
Model · Google DeepMind
Google DeepMind AI weather forecasting model producing hourly global forecasts at up to 5 km resolution.
Model · Princeton
Fully open 3B text-to-image diffusion model from Princeton with released weights, training code, and data pipeline.
Agent · Google Research
Google Research agent that turns a natural-language Earth-science question into a data-driven prediction.
Model · Baidu
Baidu's open-weights 280B-parameter sparse MoE base model (14B active) with 1M-token context, released pre-post-training.
Model · Google
Google's open-weights 26B diffusion language model that drafts blocks of text in parallel.