shiyu-coder/kronos

Kronos is an open-source AI model trained specifically to read and predict financial market data, the kind of candlestick charts (called K-lines) that traders use to track price movements over time. Think of it like a language model for stocks and crypto: instead of learning from text, it learned from price patterns across data from over 45 global exchanges. The problem it solves is that standard AI models are built for general tasks and struggle with the noisy, fast-moving nature of financial market data. Kronos was purpose-built for this challenge. It works in two stages: first, it converts raw candlestick data (open, high, low, close, and volume prices) into structured tokens, basically a compressed numeric language, then a large Transformer model (the same type of architecture that powers modern AI chatbots) reads those tokens to predict future price movements. You would use this if you are building automated trading strategies, doing quantitative research, or experimenting with AI-powered market forecasting. It can predict future candlestick values for a given asset, and it comes in several sizes (mini, small, base) so you can choose based on how much computing power you have. The models are available on Hugging Face for download. Kronos is written in Python and accepted at the AAAI 2026 research conference.