Loading...
Run Chronos, TimesFM, Moirai, and more through a single endpoint. Production-grade forecasting — no GPUs required.
—
Models
—
Online Now
—
Model Families
—
Avg Latency
Chronos, TimesFM, Moirai, Lag-Llama, MOMENT — access every major time series model through a single endpoint.
Browse all →No GPU provisioning. No model weights to manage. Send your time series data, get forecasts back in milliseconds.
View models →Point forecasts, prediction intervals, anomaly detection, imputation, and multi-horizon outputs from every model.
Compare →Autoscaling GPU infrastructure, 99.9% uptime SLA, and transparent per-token pricing for every model.
Get started →0 models across 0 families
import requests
resp = requests.post(
"https://api.tsfm.ai/v1/forecast",
headers={"Authorization": "Bearer YOUR_API_KEY"},
json={
"model": "amazon/chronos-2",
"series": [10.2, 11.5, 9.8, 12.1, 13.4],
"horizon": 12,
"quantiles": [0.1, 0.5, 0.9],
},
)
print(resp.json()["forecasts"])Time Series Foundation Models (TSFMs) are large neural networks pre-trained on billions of observations across thousands of diverse time series. Unlike classical methods that must be fit individually to each dataset, TSFMs generalize out-of-the-box — send any series and get accurate forecasts with zero training.
Prophet and ARIMA are excellent tools for well-understood, stationary series where you can invest time in per-series tuning. TSFMs shine when you need to forecast many heterogeneous series at scale with minimal setup.
No GPUs to provision. No models to train. No infrastructure to manage.
POST your historical observations to the forecast endpoint. Pick any model from the catalog.
Read the docs →Receive point forecasts, prediction intervals, and uncertainty quantiles in milliseconds.
Try the playground →Free tier included. No credit card required.