Use this file to discover all available pages before exploring further.
Alloy (Auto Model Routing) is currently in Research Preview.
Alloy (Auto Model Routing) automatically chooses the best model for each Droid task. Instead of manually picking a single model for every session, Alloy evaluates the task and routes work to the model that balances quality, latency, and cost.Alloy is available in the Factory CLI and App and works with the same enterprise controls as other models.
Alloy is designed for teams that want strong results without manually changing model selection for every workflow.
Best performance at lower cost
In Factory evaluations, Alloy maintained frontier-level performance on challenging benchmarks while reducing costs roughly 20–25% compared with always using top-tier models.
Improved model selection
Droid can route routine steps to faster or lower-cost models and reserve stronger models for work that needs deeper reasoning.
Enterprise-ready controls
Alloy respects the same org, project, and user-level model controls used across the Factory platform.
Enterprise admins can manage Alloy through the same model governance system used for other Factory-supported models, including allowing or restricting Alloy at the organization level like any other model.For more information on enterprise control of models, see Models, LLM Gateways & Integrations.
Alloy is a per-task router that uses a mix of session and per-request routing. It strongly considers prompt cache maintenance and savings in its optimization.
Yes. Alloy maintains frontier-level performance on challenging benchmarks like Terminal Bench 2 and Legacy Bench, and provides the best balance of quality, latency, and cost.
