Kronos Mini

Kronos Mini is the smallest checkpoint in the NeoQuasar Kronos family — a tokenizer-based time-series foundation model purpose-built for financial OHLCV candlestick data. At 4.1M parameters it is the most lightweight way to access the Kronos design, the entry point to the family for cost- and latency-sensitive financial forecasting.

Architecturally Kronos treats forecasting like a language-modeling problem over price action: a custom tokenizer discretizes price movements into a vocabulary, and a GPT-style decoder autoregressively predicts future candlestick patterns. The Mini variant uses a 2K-token vocabulary and a 2048-token context window. Across the family, the official Kronos model cards describe pretraining on over 12B K-line records from 45 global exchanges, which is what gives the checkpoints their broad zero-shot coverage of market data.

On TSFM.ai reach for Kronos Mini when you want financial candlestick forecasting at the cheapest footprint, or when running many instruments in parallel where per-call cost matters more than peak accuracy. Step up to Kronos Small for a larger tokenizer vocabulary and more transformer capacity on complex multi-asset tasks, or to Kronos Base for maximum forecast quality on financial workloads.