Simudyne has partnered with BMLL to power its Large Market Models (LMMs), AI foundation models built for the next generation of intraday trading, execution and risk. The models are being trained on BMLL’s data and will be released as new modules within Pulse, Simudyne’s platform for simulating capital markets.
Language models learned from vast amounts of text. Market models need to learn how order books behave: how liquidity forms, how it disappears, and how markets react under pressure. That requires data at the deepest level of market activity, harmonised across venues, so a model learns consistent patterns rather than the quirks of individual feeds.
That is why the LMMs are trained on BMLL’s Level 2 and Level 3 data, from aggregated depth down to every individual order placed, cancelled and re-priced. Learning at this depth lets the models capture how markets behave, not just predict the next price.
Pulse already simulates the order book as a system of agents that respond to market conditions, including the orders you submit. The LMMs extend it with new modules, including PulseGPT and PulseFlow, which bring participant behaviour learned directly from market data into reactive market sandboxes. Algo traders can test execution strategies against more realistic market responses. Exchanges can stress-test matching engine logic and rule changes before they reach production. Risk and research teams can explore plausible scenarios beyond the historical record.
The work is enabled by the BMLL Activate Data Credits Programme, which helps partners build new analytics and data-driven solutions on BMLL data.
Pulse is available to try now at pulse.simudyne.com. To hear more about the LMMs and upcoming modules, contact info@simudyne.com.