Moirai-1.0-R-Small

Moirai-1.0-R-Small is the smallest dense checkpoint in Salesforce's original Moirai family, the first-generation universal forecasting line that also ships as Base and Large. At 14M parameters it is the most cost-efficient way to access the Moirai design, trading raw capacity for the lowest serving footprint in the dense 1.0-R lineup.

Architecturally it is a masked-encoder transformer with multi-patch projections, any-variate attention, and a mixture-distribution output head. The any-variate attention lets it reason over arbitrary numbers of target variables and dynamic covariates rather than being fixed to a single series, and the mixture output head produces probabilistic forecasts with calibrated quantiles instead of a bare point estimate. It is pretrained on the LOTSA corpus, the large open archive assembled for the Moirai work, which is what gives the checkpoint its broad zero-shot reach across domains.

On TSFM.ai reach for it when you need genuine multivariate, covariate-aware, probabilistic forecasting at the cheapest possible cost, or when running many series in parallel where per-call latency matters more than peak accuracy. Step up to Moirai-1.0-R-Base for stronger accuracy on heterogeneous multivariate settings, or to Moirai-1.0-R-Large for the first-generation architecture at its highest published capacity; consider the newer 1.1-R checkpoints if low-frequency yearly and quarterly series dominate your workload.