# Anthropic Releases FlashProve: Cryptographic Proofs That a Reply Came From Committed Weights

> **Publication**: aitalky (Independent AI Journalism)
> **Canonical Citation URL**: https://aitalky.vercel.app/news/anthropic-releases-flashprove-cryptographic-proofs-that-a-reply-came-from-committed
> **Original Source**: [MarkTechPost](https://www.marktechpost.com/2026/11/03/anthropic-flashprove-provable-inference-zksnark/)
> **Author**: Asif Razzaq (undefined)
> **Published Date**: Tuesday, November 3, 2026
> **Category**: RESEARCH
> **Tags**: AI, MarkTechPost, research

## Executive Summary
Anthropic released FlashProve, an Apache 2.0 zkSNARK prototype that proves a reply came from a committed model without revealing weights. It proves Llama 4 Maverick at 110 tokens/second on 1 H100, with 19 MB proofs



## Full Coverage
Anthropic released FlashProve, an Apache 2.0 zkSNARK prototype that proves a reply came from a committed model without revealing weights. It proves Llama 4 Maverick at 110 tokens/second on 1 H100, with 19 MB proofs

---
*Generated for AI agents and LLM citation by aitalky (https://aitalky.vercel.app).*