Radar / AI models / Kronos

Kronos: an open foundation model that treats market candlesticks as language

Kronos is a decoder-only transformer family from Tsinghua researchers (accepted at AAAI 2026, MIT-licensed) purpose-built for financial K-line data. A specialized tokenizer converts continuous OHLCV candlesticks into discrete tokens and the model pretrains autoregressively on over 12 billion K-line records from 45+ global exchanges across seven time granularities. Open variants range from Kronos-mini (4.1M) to Kronos-base (102.3M); reported results include +93% price-forecasting RankIC over leading general time-series foundation models, 9% lower volatility-forecast MAE, and synthetic K-line generation fidelity up 22%.

trendingmodels · decisionJEV traction 0.37shiyu-coder/Kronos · 43,541 ★added 2026-10-20

Open Kronos →View on the radar

Why it matters

The financial-markets entry in the new wave of vertical foundation models: instead of treating prices as generic sequences, its tokenizer natively encodes candlestick structure (open/high/low/close relationships, volume, session gaps), so the transformer learns temporal patterns rather than data representation.

What you could build with it

A two-person quant research team could fine-tune Kronos-mini on a single niche futures market and generate synthetic K-line histories to stress-test their strategies against market regimes that have never occurred in the real data.

Does it hold up?

Strong for research, not for live trading: the official live BTC demo and open weights make it runnable today, and third-party walkthroughs confirm quick setup - but the RankIC gains are paper benchmarks, not live P&L, and regime changes can break any pretrained market model.

Built with Kronos

Kronos: A Domain-Specific Foundation Model for the Language of Financial Marketstsfm.ai · Technical deep dive placing Kronos in the 2026 wave of domain-specialist time-series models (Toto for observability, Cisco's telemetry model); notes the three open variants are hosted directly and that it crossed 25K GitHub stars within months of the paper.
Vertical Models Just Stole the Inference Stackdev.to · Fresh analysis framing Kronos (41.8K stars, still adding ~355/day) alongside Timestral and Flare as proof that the vertical-model paradigm is winning where general time-series foundation models underperform.
Kronos - AI learns the language of marketscloud9markets.com · Practical 60-second guide: live BTC/USDT 24-hour forecast demo, clone-and-install walkthrough, and an explicit reality check that it is a research model for study, not a signal generator to outsource risk to.

Learn more

technical deep dive →

First spotted on github: source.

More AI models

OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72

Get the week's best AI launches, plus 3 ideas worth building

One email every Saturday. Ranked by traction, not hype. Free.