Salesforce Moirai — hosted and ready
Run Moirai 1.1 through the TSFM.ai API. A universal time series forecasting model with covariate support, multiple sizes, and context lengths up to 5000 — no infrastructure to manage.
{
"model": "salesforce/moirai-1.1-r-base",
"inputs": [{
"item_id": "store-revenue",
"target": [8500, 8700, 9100, 8900, 9300, 9500],
"start": "2026-03-01T00:00:00Z"
}],
"parameters": {
"prediction_length": 21,
"freq": "D",
"quantiles": [0.1, 0.5, 0.9]
}
}Same request shape as every other model on TSFM.ai.
Architecture
Universal transformer with any-variate attention
Context
Up to 5000 time steps — longest in the catalog
Covariates
Supports exogenous variables alongside the target series
Why teams choose Moirai
Universal forecasting
Moirai is trained across diverse domains and frequencies. It generalizes well to new datasets without fine-tuning, making it a strong default for teams with heterogeneous data.
Covariate support
Pass exogenous variables — promotions, holidays, weather — alongside your target series. Moirai's any-variate attention mechanism incorporates them directly into the forecast.
Longest context in the catalog
With support for up to 5000 time steps of history, Moirai can capture long-range seasonal patterns that shorter-context models miss. Ideal for weekly and monthly data with yearly cycles.
Compare modelsMoirai variants on TSFM.ai
Moirai 1.1-R ships in multiple sizes. All variants support covariates and extended context lengths.
Getting started with Moirai
- 1
Pick a variant
Start with moirai-1.1-r-base for most workloads. Use small for high-throughput batch jobs or large when forecast accuracy is the top priority.
- 2
Send a forecast request
POST to /v1/forecast with model set to your chosen variant. Include covariates in the inputs array if you have exogenous signals like promotions or weather.
- 3
Leverage the extended context
If your data has strong yearly seasonality, pass up to 5000 historical observations. More context helps Moirai capture long-range patterns that improve forecast quality.
Frequently Asked Questions
Try Moirai on your data
Send a series with covariates, pick a horizon, and see how Moirai 1.1 performs. No setup, no weights, no GPU — just an API key.