ResearchNews
Anthropic's FlashProve prototype proves which model wrote a reply
Anthropic's FlashProve research prototype generates cryptographic proofs that a reply came from a specific model, without revealing weights. Anthropic reports about 110 tokens a second on one H100.
