shiyu-coder/Kronos

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

39500 stars · 6578 forks observed · 2026-09-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-09-28, confidence 0.97

Kronos is an open‑source decoder‑only foundation model trained on financial candlestick (K‑line) data from dozens of global exchanges. It provides a tokenizer that converts OHLCV time‑series into hierarchical discrete tokens and a large autoregressive transformer for downstream quantitative tasks.

Use cases

• Generate synthetic financial time‑series for back‑testing trading strategies
• Fine‑tune the model for price prediction or anomaly detection on specific market data
• Extract patterns or embeddings from raw K‑line data for clustering or classification
• Create a unified model that can handle multiple quantitative finance tasks
• Develop a research pipeline for studying market micro‑structure using language‑model techniques

When to choose

• You need a pre‑trained model that understands the structure of financial candlestick data
• You want to leverage transformer‑based language modeling techniques for quantitative finance
• You plan to fine‑tune a model on your own market data without building a tokenizer from scratch

When to avoid

• Your application requires real‑time inference on low‑power edge devices
• You need support for non‑financial time‑series data without adapting the tokenizer
• You prefer a model with a permissive license other than MIT

Facets

library · maturity active

llm-inference machine-learning nlp artificial-intelligence large-language-models data-science python windows financial-markets candlestick k-line foundation-model transformer tokenizer tokenization linux macos

Member repositories

For agents

markdown · JSON · MCP: product_card(name="shiyu-coder/Kronos")