Amazon Models

Amazon Chronos models — hosted and ready

Run every Chronos and Chronos-Bolt variant through one API. Choose from mini to large, get probabilistic forecasts with quantiles, and swap sizes without changing your integration code.

Chronos & Chronos-BoltMini through LargeProbabilistic forecasts
Chronos-BoltPOST /v1/forecast
{
  "model": "amazon/chronos-bolt-base",
  "inputs": [{
    "item_id": "web-traffic",
    "target": [1200, 1350, 1100, 1420, 1380, 1500],
    "start": "2026-03-01T00:00:00Z"
  }],
  "parameters": {
    "prediction_length": 14,
    "freq": "D",
    "quantiles": [0.1, 0.5, 0.9]
  }
}

Swap amazon/chronos-bolt-base for any Chronos variant — same request shape.

Family

Chronos (original) and Chronos-Bolt (optimized) variants

Sizes

Mini, Small, Base, and Large — trade latency for accuracy

Architecture

T5-based encoder-decoder with quantile token generation

Amazon model variants on TSFM.ai

All variants are hosted and available through the same API endpoint. Choose the size that fits your latency and accuracy requirements.

Why teams choose Chronos

Multiple sizes for different workloads

Start with mini for high-volume batch scoring, use base for production defaults, and reserve large for accuracy-critical forecasts. One request shape across all sizes.

Native probabilistic forecasts

Chronos generates quantile tokens during inference, giving you prediction intervals without post-hoc calibration. Use them for safety stock, capacity buffers, and risk thresholds.

Chronos-Bolt for lower latency

Bolt variants use a more efficient architecture that cuts latency by 2-5x compared to original Chronos at comparable accuracy. Ideal for interactive and streaming workloads.

Compare with TimesFM

Getting started with Chronos

  1. 1

    Pick a variant

    Start with chronos-bolt-base if you are unsure. It offers the best balance of latency and accuracy for most production workloads.

  2. 2

    Send a forecast request

    POST to /v1/forecast with model set to your chosen variant. The request shape is identical across all Chronos models.

  3. 3

    Compare sizes on your data

    Run the same series through mini, base, and large. Compare forecast accuracy and latency to find the optimal size for your use case.

Frequently Asked Questions

Try Chronos on your data

Pick a variant, send a series, and get back probabilistic forecasts. No setup, no weights, no GPU — just an API key.