IBM releases SOTA Granite Time Series PatchTST-FM-r2 model with commercial-friendly license
IBM has released the Granite Time Series PatchTST-FM-r2 model, a foundation model specifically optimized for time-series forecasting. The model is distributed under the Apache 2.0 license, enabling unrestricted commercial use.
Verified State Diff
Impact & Verification Analysis
Data scientists, financial analysts, supply chain engineers, and enterprise developers building predictive forecasting systems.
This release democratizes access to state-of-the-art (SOTA) time-series foundation models, reducing the computational overhead for enterprises to implement high-accuracy forecasting without needing to train models from the ground up.
Full Fact Overview
The Granite Time Series PatchTST-FM-r2 model utilizes the PatchTST architecture, which segments time-series data into patches to capture local semantic information and long-term dependencies more effectively than point-based transformers. By releasing this as a foundation model (FM), IBM provides a pre-trained architecture capable of zero-shot or fine-tuned forecasting across diverse temporal datasets. The shift to an Apache 2.0 license removes the restrictive barriers often associated with proprietary research models, allowing enterprise integration into production-grade predictive pipelines.