What is ModelRoute?
ModelRoute is an AI model execution marketplace. You submit a request with a model slug and canonical input. ModelRoute routes it to the best available provider, executes it, and returns a canonical response. You never see or interact with the underlying provider.Why use ModelRoute?
Provider Abstraction
One API for hundreds of models. ModelRoute handles provider selection, authentication, and format translation.
Normalized Errors
Every provider error is classified into 12 deterministic error codes. No more parsing provider-specific error formats.
File Isolation
Files are uploaded to ModelRoute, not to providers. Providers never see your raw files — ModelRoute handles secure transfer.
Government-Grade Reliability
Built for mission-critical workloads. Automatic failover, retry logic, and multi-provider redundancy.
How it works
Send a request
POST your execution request with a model slug and input. ModelRoute validates, estimates cost, and places a balance hold.
ModelRoute routes and executes
The platform selects the best available provider based on cost, latency, and availability. Your request is translated into the provider’s format and executed.
Key concepts
| Concept | Description |
|---|---|
| Execution | A single AI model run. Has a lifecycle: PENDING, PROCESSING, COMPLETED, FAILED. |
| Model slug | A canonical identifier for a model (e.g., flux-1.1-pro). Provider-agnostic. |
| File reference | A file_<uuid> identifier for uploaded files. Used in execution inputs and outputs. |
| Balance hold | A cost estimate locked before execution. Settled to actual cost on completion, released on failure. |
| Idempotency key | A unique key to prevent duplicate executions. Same key = same result. |
Next steps
Quickstart
Run your first execution in 5 minutes.
Authentication
Learn how API keys work and how to manage them.