How an inference provider can prove they're not serving a quantized model

End-to-end private LLM inference