GPU Router
Loading models…

Fetching the model catalog.

DEFERRED INFERENCE

Batches

Set a target discount, give the market time, and collect your results.

Execute only when the price qualifies

Both input and output rates must meet your target. Current live adapters cannot guarantee those caps, so jobs wait and may expire. Operators must enable inference and encrypted storage. No credits are reserved while waiting.

Queue requests

Kept in memory for this page only. Changing networks clears it.Keep this unchanged when retrying. Use a new key for a separate batch.

Up to 50 paid text requests. Privacy redaction is enabled in the example; review its limits before sending sensitive data.

Track a batch

Submit a batch or paste an existing ID to view its requests, cancel waiting work, and download results.

Connect to a gateway with a matching protocol identity before using your API key.