Moirai 2.0: When Less Is More for Time Series Forecasting

TL;DR

Moirai 2.0 is a streamlined, decoder-only time series model that improves probabilistic forecasting accuracy and efficiency, outperforming larger models and previous versions while maintaining a favorable speed and size trade-off.

Contribution

It introduces a simplified decoder-only architecture with quantile forecasting and multi-token prediction, achieving state-of-the-art results with enhanced efficiency and robustness.

Findings

Moirai 2.0 ranks among top models on Gift-Eval benchmark.

It is twice as fast and thirty times smaller than Moirai 1.0-Large.

Performance plateaus with more parameters and declines at longer horizons.

Abstract

We introduce Moirai 2.0, a decoder-only time-series foundation model trained on a new corpus of 36M series. The model adopts quantile forecasting and multi-token prediction, improving both probabilistic accuracy and inference efficiency. On the Gift-Eval benchmark, it ranks among the top pretrained models while achieving a strong trade-off between accuracy, speed, and model size. Compared to Moirai 1.0, Moirai 2.0 replaces masked-encoder training, multi-patch inputs, and mixture-distribution outputs with a simpler decoder-only architecture, single patch, and quantile loss. Ablation studies isolate these changes -- showing that the decoder-only backbone along with recursive multi-quantile decoding contribute most to the gains. Additional experiments show that Moirai 2.0 outperforms larger models from the same family and exhibits robust domain-level results. In terms of efficiency and model size, Moirai 2.0 is twice as fast and thirty times smaller than its prior best version, Moirai 1.0-Large, while also performing better. Model performance plateaus with increasing parameter count and declines at longer horizons, motivating future work on data scaling and long-horizon modeling. We release code and evaluation details to support further research.