Time Series Foundation Models in Financial Markets
Pre-trained forecasting models top general benchmarks, but financial returns are a different beast. What Chronos-2, TimesFM and their peers can and cannot do for trading desks in 2026.
AI & Quant Models
Time series foundation models, deep learning on order books, reinforcement learning for market making, ML in HFT and LLM trading agents.
AI & Quant Models: an overview
Machine learning is now part of most quantitative trading stacks, but microstructure data is noisy, non-stationary and latency constrained.
We separate what works in production from what works only in papers, from foundation models and DeepLOB to reinforcement learning and LLM research agents.
Latest AI & Quant Models research
Browse all research Key concepts
- Foundation model
- A large model pre-trained on broad data and adapted to specific forecasting tasks.
- Alpha decay
- The decline in a signal’s predictive power as markets adapt and competitors copy it.
- Inference latency
- Time a model takes to produce a prediction, critical in fast strategies.
Research roadmap
What we cover in AI & Quant Models
Our editorial plan for this section. Published guides are linked; the rest are in research.
- Live Time Series Foundation Models in Financial Markets
- Soon Deep Learning on Limit Order Book Data
- Soon Machine Learning in High-Frequency Trading
- Soon Reinforcement Learning Applied to Market Making
- Soon Large Language Models in Trading: Agents and Applications
- Soon Transformers in Trading and Financial Forecasting
- Soon Alpha Decay and Strategy Crowding