Show HN: T0: We trained a 256M SOTA open-weights time-series forecasting model

Category: ai-ml

Tags: time-series, forecasting, foundation-model, pytorch, mlx

Score: 7.0/10 (Innovation: 7, Technical: 8, Documentation: 7, Utility: 6)

T0 is a 256M-parameter open-weights foundation model for time-series forecasting that supports probabilistic forecasts with historical and known-future covariates. It provides a PyTorch/MLX runtime, batched inference, and preprocessing options for missing data, making state-of-the-art forecasting accessible to practitioners.

Target audience: data scientists, machine learning engineers, researchers

Repository: https://github.com/theforecastingcompany/tfc-t0 · Jupyter Notebook · Apache-2.0 · 67 stars

View on Hacker News