cost. The recommended production route is Amazon SageMaker, which is billed by AWS.
Who it is forA team deciding what to run on a CPU box or a small GPU.
The problemRunning the largest model everywhere wastes time and memory.
The outcomeA model choice you can justify, with the parameter count written down.
What you need first
- Python with pip and `pip install chronos-forecasting`
- pandas with pyarrow for the DataFrame examples
- A CUDA GPU for the README example, which sets device_map to cuda
Steps
- Read the Available Models table, which lists IDs from amazon/chronos-bolt-tiny at 9M up to amazon/chronos-t5-large at 710M
- Note that the README describes Chronos-Bolt as up to 250 times faster and 20 times more memory efficient than the original Chronos models of the same size
- Install the package with
pip install chronos-forecastingand try the smallest candidate first - For production the README recommends deploying Chronos-2 to Amazon SageMaker through AutoGluon-Cloud or SageMaker JumpStart
Prompt for your agent
Using amazon-science/chronos-forecasting, compare the listed Chronos-2, Chronos-Bolt and Chronos-T5 model IDs for <my forecasting workload> and recommend one with the reasoning.
Caveat. The benchmark claims in the README are the authors' own.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for amazon-science/chronos-forecasting and "Choose the Chronos model that fits your budget"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer