Moirai-MoE-1.0-R-Small

Moirai-MoE-1.0-R-Small is the lightweight checkpoint in Salesforce's sparse-expert Moirai-MoE family, the smaller counterpart to Moirai-MoE-1.0-R-Base. With around 0.47B stored parameters it is the most cost-efficient way to access the Moirai-MoE design, which departs from the dense 1.0-R line by routing tokens through specialized experts instead of one monolithic network.

Architecturally it is a sparse mixture-of-experts decoder-only transformer with probabilistic output heads, sharing the same MoE routing as the Base variant but at a smaller stored size. The sparse routing activates only a subset of experts per token, which is what makes the family parameter-efficient and high-throughput at inference, while the probabilistic heads supply multivariate forecasts with calibrated quantiles. The checkpoint card is sparse; Salesforce's official Moirai-MoE materials describe large heterogeneous time-series pretraining for the setup rather than a separate narrow corpus for this exact checkpoint.

On TSFM.ai reach for it when you want the Moirai-MoE design at its cheapest, or for high-throughput zero-shot multivariate forecasting where per-call cost matters. Step up to Moirai-MoE-1.0-R-Base for the family's stronger zero-shot accuracy. Note that this checkpoint is released under a non-commercial license (CC-BY-NC-4.0), so prefer a dense Moirai checkpoint for commercial workloads.