Insider Brief
- FirstQFM reported that its quantum reservoir computing system, running on Rigetti hardware, outperformed the classical models evaluated in financial forecasts covering 43 anonymized assets and 600 out-of-sample periods.
- The system achieved 52.02% directional accuracy and 112% higher cumulative gross portfolio profit than the strongest classical benchmark by profit, with an annualized Sharpe ratio of 1.22.
- The evaluation used data anonymized by SW7 Group and chronological training, validation and test periods, with model selection based only on validation data.
- Photo by Adam Śmigielski on Unsplash
PRESS RELEASE — FirstQFM today announced results from a financial forecasting evaluation in which its proprietary foundation-model-based quantum reservoir computing (QRC) system, running on Rigetti Computing ’s Cepheus -1-108Q system using data supplied by SW7 Group, outperformed state-of-the-art classical benchmarks across directional accuracy, macro-F1, gross portfolio P&L, and annualized Sharpe ratio.
The evaluation covered next-day forecasts across 43 anonymized assets and 600 out-of-sample periods. FirstQFM ’s foundation-model-based QRC system achieved 52.02% directional accuracy and 51.76% macro-F1, the highest results among the models evaluated.
In an equally weighted portfolio of all 43 assets, FirstQFM ’s system delivered 112% higher cumulative gross P&L than the strongest classical benchmark by P&L with an annualized Sharpe ratio of 1.22 versus the highest classical benchmark Sharpe of 1.02. Across individual assets, the system achieved directional accuracy above 50% for 36 of 43 assets and positive gross P&L for 41 of 43.
Dr. Subodh Kulkarni, Chief Executive Officer, Rigetti Computing : “Empirical work such as FirstQFM ’s helps us probe the capabilities and limits of today’s quantum computers, while helping hardware teams understand which improvements matter most. As we advance towards fault tolerance, continued investment in algorithms, applications, and the tools and skills needed to use them will be essential to turn increasingly capable machines into practical value.”
The comparison included classical machine learning models, statistical models, and time series foundation models, including Chronos-2 (Amazon), Granite TTM-R2 (IBM), and Toto 2.0 (Datadog).
SW7 Group anonymized the assets and forecasting task before providing the data to FirstQFM . The evaluation used chronological training, validation, and test periods, with model selection based only on validation data. FirstQFM’s quantum features were generated using Rigetti ’s quantum hardware.
A five-asset portfolio using FirstQFM ’s system, selected through a historical diversification methodology, delivered higher directional accuracy, macro-F1, and cumulative gross P&L than the full 43-asset portfolio, with an annualized Sharpe ratio of 1.73.
The detailed methodology and results are available in FirstQFM ’s accompanying technical blog: https://firstqfm.com/blog/financial-forecasting-qrc-cepheus-1-108q.
Oleksiy Kondratyev, Head of Risk, SW7 Group: “We are focusing on the development of innovative approaches to solving challenging quant finance problems. The experimental results clearly demonstrate that QRC offers us an opportunity to build viable QML solutions with the potential to outperform classical benchmarks.”
FirstQFM released the beta version of its foundation-model-based QRC forecasting system in July 2026. Built on proprietary foundation model technology, the system operates at a billion-parameter scale and is developed for applications across verticals, including asset management, banking, healthcare, predictive maintenance, and energy.
Vish Ramakrishnan, CEO, FirstQFM AB: “FirstQFM’s foundation model based QRC solution leverages our patented proprietary foundation models to provide outperformance over classical time series models and delivers a utility-scale, commercially ready solution with demonstrated value for enterprise customers.”