What is Kronos
Kronos is the first K-line chart foundation model for the financial market, jointly open-sourced by Tsinghua University and Microsoft Research Asia. It predicts future price trends by analyzing K-line data of assets such as stocks and cryptocurrencies, including opening price, highest price, lowest price, closing price, and trading volume. Kronos adopts a two-stage processing framework: an intelligent tokenizer converts continuous K-line data into discrete "financial vocabulary," and a prediction large model based on the Transformer architecture learns patterns from historical data to predict future trends. The model's training data covers 45+ exchanges worldwide, enabling it to adapt to the high volatility and noise of financial data.
Kronos offers multiple parameter models, ranging from 4.1M to 499.2M parameters, to meet different needs. It can be loaded with just 4 lines of code, and automatically outputs prediction results after inputting historical data. Kronos provides a real-time BTC/USDT prediction dashboard, integrates Qlib for backtesting validation, and supports A-share market data. On benchmark datasets, Kronos improves RankIC for price series forecasting by 93% over leading TSFM and by 87% over the best non-pretrained baseline. In volatility forecasting, it achieves a 9% lower MAE, and in the generation fidelity of synthetic K-line sequences, it achieves a 22% improvement.
Kronos's Main Features
-
K-line Chart Interpretation: Kronos can analyze K-line data of financial assets such as stocks and cryptocurrencies, including opening price, highest price, lowest price, closing price, and trading volume, to predict future price trends.
-
Two-Stage Processing Framework: It uses an intelligent tokenizer to convert continuous K-line data into discrete "financial vocabulary," and then uses a prediction large model based on the Transformer architecture to learn patterns from historical data for prediction.
-
Multiple Model Choices: It provides multiple pretrained models ranging from 4.1M to 499.2M parameters, meeting different computational and application needs.
-
Simple and Easy to Use: It can load the model with just 4 lines of code, and automatically outputs prediction results after inputting historical K-line data.
-
Real-Time Prediction Demo: It provides a real-time BTC/USDT prediction dashboard, displaying future trends based on the model's calculation results.
-
Supports the A-Share Market: It integrates Qlib for backtesting validation, and also provides a complete fine-tuning pipeline, which can be adapted to your own trading strategy.
-
High-Performance Prediction: On benchmark datasets, Kronos improves RankIC for price series forecasting by 93% over leading TSFM and by 87% over the best non-pretrained baseline.
Kronos's Technical Principles
-
Two-Stage Processing Framework: Kronos adopts a two-stage processing framework. First, an intelligent tokenizer converts continuous K-line data into discrete "financial vocabulary," and then a prediction large model based on the Transformer architecture learns patterns from historical data to predict future trends.
-
Intelligent Tokenizer: The tokenizer extracts key information from K-line data and converts it into discrete symbols that the model can understand and process, providing the basis for subsequent prediction.
-
Transformer Architecture: The prediction large model is based on the Transformer architecture, which can process long sequence data and capture long-term dependencies in time series, thereby predicting future price trends more accurately.
-
Pretraining and Fine-Tuning: Kronos provides a variety of pre-trained models that users can fine-tune according to specific tasks, making them better adapt to specific financial market data and forecasting needs.
-
Multi-source data training: Model training data covers 45+ exchanges worldwide, enabling it to adapt to the characteristics of financial data in different market environments and providing strong generalization capability.
-
Time series modeling: Kronos focuses on time series modeling, is able to handle the high volatility and noise of financial data, and extract valuable information for forecasting.
Kronos project address
- Github repository：https://github.com/shiyu-coder/Kronos
- arXiv technical paper：https://arxiv.org/pdf/2508.02739
Kronos application scenarios
-
Stock market forecasting: Kronos can analyze K-line data of the stock market, forecast the future trend of stock prices, and help investors make wiser investment decisions.
-
Cryptocurrency trading: Suitable for the cryptocurrency market, by analyzing cryptocurrency price fluctuations, it provides price forecasts for traders and assists in formulating trading strategies.
-
Quantitative trading strategy development: Kronos can serve as an important tool in quantitative trading strategy development, providing market trend forecasts for strategies and optimizing the generation of trading signals.
-
Market sentiment analysis: By interpreting K-line charts, Kronos can reflect changes in market sentiment, help investors understand the emotional tendencies of market participants, and grasp the pulse of the market.
-
Risk management: Using its forecasting capability, investors can better assess investment risks, take risk control measures in advance, and reduce potential losses.
-
Financial data research: Provides financial researchers with a powerful tool for studying issues such as the patterns of financial markets, price formation mechanisms, and market efficiency.