Chronos-Bolt (Tiny)

Chronos-Bolt Tiny is the smallest checkpoint in Amazon's Chronos-Bolt family, at 9M parameters. It is the cost and latency floor of the line: a probabilistic zero-shot forecaster small enough to favor throughput and price over peak accuracy, aimed squarely at latency-sensitive and high-volume work.

Architecturally it keeps the T5-style encoder-decoder backbone shared across Chronos-Bolt, but swaps token-by-token rollout for direct multi-step quantile prediction over patchified history. Predicting the whole horizon in one pass is what makes Bolt materially faster and lighter than the original autoregressive Chronos line. It is pretrained on nearly 100B time-series observations drawn from large public corpora and synthetic pretraining data, as documented by the official Chronos-Bolt model cards.

On TSFM.ai, reach for Tiny in latency-sensitive applications, edge-style deployments, and high-volume batch forecasting where you need quantile forecasts at minimal cost per series. Step up to Chronos-Bolt Mini or Small for a better accuracy-latency balance, to Chronos-Bolt Base when you want the family's higher quality ceiling, or to Chronos-2 when you need covariates or multivariate cross-learning rather than fast univariate throughput.