#3131st of 52 in Open modelsFor data and documents, data analysis
chronos-forecasting
Pretrained zero-shot time series forecasting models with a Python pipeline
- Stars
- 6.2k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- Sep 2026
- Language
- Python
5,017 on 3 Apr 20266,230 on 30 Oct 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Python package that loads the Chronos family of pretrained time series models (Chronos-2, Chronos-Bolt, original T5-based Chronos) from Hugging Face and produces probabilistic quantile forecasts. Chronos-2 handles univariate, multivariate and covariate-informed forecasting zero-shot and takes pandas DataFrames through predict_df. Model sizes range from 8M to 710M parameters.
Who it is for: Data scientists and ML engineers needing zero-shot time series forecasts
Strengths
- Zero-shot forecasting with covariates and multivariate series in Chronos-2 (120M parameters)
- Size range of 8M to 710M parameters across three model generations
- DataFrame-based predict_df API returns quantile forecasts; installs with pip
- Apache-2.0, with notebooks and AWS deployment guides (SageMaker JumpStart, AutoGluon-Cloud)
Weaknesses
- README gives no RAM or VRAM requirements
- Production deployment guidance centers on AWS
- No hosted demo; quick start runs in Colab or SageMaker Studio Lab notebooks
What it needs
- GPU optional
- Needs Hugging Face model hub, pandas[pyarrow]
- Models: Chronos-2, Chronos-Bolt, Chronos-T5
Quick answers
Is chronos-forecasting open source?
Yes. Its code is on GitHub under the Apache-2.0 license.
Does chronos-forecasting need a GPU?
No. It runs on a CPU and can use a GPU when you have one.
Also in Open models
See all 52| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 95 out of 100 |
| 2 | llmfitRust tool that checks which LLMs fit your RAM, CPU and GPU | 87 out of 100 |
| 3 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 86 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 77 out of 100 |
| #5 | yolov5PyTorch YOLOv5 object detection, segmentation and classification models with training code | 75 out of 100 |