New Inferock

What Inferock does is “accountable reasoning”: every time a large model is called, it keeps track of which route is used, where the execution takes place, how much cost is incurred, and why any failures occur. All these details are recorded in a verifiable manner, and end-to-end encrypted channels for confidential reasoning are also provided.

There are three main uses: 1. **Inferock Watch**: You can connect it to your own model provider’s account for performance, cost, and failure monitoring. Using your own API key is completely free. 2. **Inferock Serve**: This is a serverless inference service that operates on a prepaid basis. There is no need to use separate API keys from various providers, and the service will provide compensation in case of failures that meet certain criteria. 3. **Inferock Dedicated**: This option offers exclusive computing resources for your specific needs.

It is primarily aimed at development teams and enterprises that have specific requirements regarding data compliance and call auditing; individuals who simply need to adjust models on a casual basis may not find it very useful.