BIS launches BISTRO
The Bank for International Settlements launched BISTRO—a ChatGPT-style tool for time series that offers macro forecasts and scenario exploration with code and paper available on GitHub (x.com). It’s positioned as a practical environment for experimenting with causal inference, local projections, and macro forecasting models (x.com).
BIS published Working Paper No. 1337, titled "Introducing BISTRO: a foundational model for unconditional and conditional forecasting of macroeconomic time series," on 16 March 2026. (bis.org) The BIS Quarterly Review (March 2026) frames BISTRO as a time‑series model built on transformer architectures and trained on the BIS’s macroeconomic archives. (bis.org) The authors listed on the paper are Batuhan Koyuncu, Byeungchun Kwon, Marco Jacopo Lombardi, Fernando Perez‑Cruz and Hyun Song Shin. (bis.org) BIS says the model was fine‑tuned on thousands of macroeconomic series drawn from the BIS data portal, which provides regularly updated series such as central bank policy rates and bilateral exchange rates. (bis.org) The public repository tied to the release notes the presence of a dependency with a known critical security vulnerability in the project metadata, and the codebase includes example scripts and notebooks for replication. (github.com) Trade press and regional outlets summarized BISTRO as positioning BIS to offer a lower‑cost, flexible forecasting tool for central banks and policymakers focused on scenario analysis. (regulationasia.com)
