Together AI launches a US serverless inference endpoint for seven open models

Together AI announced a US-specific serverless inference endpoint for seven open models, saying requests sent there are processed entirely in the United States. Developers can switch the base URL while keeping their API keys and model IDs. Its docs say US-only inference costs 25% more than standard serverless rates. The processing guarantee does not cover data storage, logs or network routing, and unavailable models return an error instead of running elsewhere.