• Picks
  • Discover
  • Alternatives
  • About

© 2026 Willet

  • Newsletter
  • About
  • FAQ
  • Terms
  • Privacy & Cookies
Join the Discord (opens in a new tab)
ShiYu avatar

Kronos

Open foundation model for forecasting financial candlestick data

  • AI models and runtimes
  • Finance
View on GitHub

What it is

Kronos consists of several decoder-only transformer models, pre-trained on K-line (OHLCV) sequences gathered from over 45 exchanges. A dedicated tokenizer turns price and volume bars into discrete tokens, and a Python predictor class produces probabilistic forecasts of future candles, including batch prediction across many assets. Quant researchers and traders can load the small, mini, or base weights from Hugging Face and fine-tune them on their own market data with the included scripts.

Images

Facts

Stars
42,133 GitHub stars.
42,133

Forks
6,983
Commits
76
Contributors
18
Language
Python
License
MIT
Last commit
Apr 13, 2026
On GitHub since
July 2025
Platforms
library
Self-hostable
No
Alternative to
TimeGPT
Author
ShiYu
Tags
  • open-weights
  • machine-learning
  • trading
  • time-series
  • fine-tuning

Last synced from GitHub on Oct 20, 2026

The 10 best overlooked open-source projects, every Monday.

Share and more

Comments

No comments yet. Start the conversation.

Similar projects

  • Microsoft avatar

    Qlib

    Python platform for AI-driven quantitative investment research

    49.6k
  • Google Research avatar

    TimesFM

    Pretrained foundation model for zero-shot time-series forecasting

    34.7k
  • Prior Labs avatar

    TabPFN

    Pretrained transformer for classification and regression on tables

    8.2k