Kronos

The first open-source foundation model for financial K-line data

Kronos is the first open-source foundation model purpose-built for financial candlestick (K-line / OHLCV) data, accepted at AAAI 2026. Instead of treating price series like text or images, Kronos uses a custom two-stage architecture: a specialized tokenizer that converts continuous OHLCV data into discrete tokens, followed by an autoregressive Transformer trained on data from 45+ global exchanges. Four model sizes range from 4.1M to 499M parameters, all released under MIT license. The model learns the statistical structure of market microstructure directly from raw candlestick sequences, enabling zero-shot and few-shot forecasting across asset classes — equities, crypto, and commodities. It ships with a live BTC/USDT prediction demo, Qlib integration for A-Share markets, and a backtesting framework so researchers can evaluate strategies end-to-end. With 13.6k GitHub stars in a niche domain, the community reception has been unusually strong. Kronos matters because most "AI for trading" projects glue LLMs to news sentiment or financial reports — pattern-matching on text rather than market structure. Kronos is the rare project that treats price action itself as the primary modality, giving quants and ML researchers a base model they can fine-tune on proprietary data rather than starting from scratch on every new dataset.

Panel Reviews

The Builder

Developer Perspective

Ship

“Finally a foundation model that speaks OHLCV natively instead of forcing price data through text embeddings. The Qlib integration and Hugging Face weights mean you can fine-tune on your own tick data in an afternoon. MIT license and four model sizes give you real options.”

The Skeptic

Reality Check

Skip

“The disclaimer that this is 'not a production trading system' is doing a lot of work. Financial time series are notoriously non-stationary, and a model pre-trained on historical patterns from 45 exchanges may carry regime-specific biases that hurt live trading. Benchmark numbers on held-out historical data say nothing about alpha in live markets.”

The Futurist

Big Picture

Ship

“This is the ImageNet moment for market microstructure modeling. Once researchers have a shared pre-trained foundation to build on, progress will compound rapidly — we'll see specialized variants for volatility forecasting, options pricing, and market-making within months. AAAI acceptance gives it the academic credibility to attract serious contributors.”

The Creator

Content & Design

Skip

“If you're not deep in quantitative finance, the barrier to actually using Kronos is steep — you need to understand OHLCV data, Qlib configuration, and backtesting pipelines before you see any value. The live BTC demo is cool to watch but hard to translate into a personal use case.”