# Anthropic releases FlashProve, an open-source prototype for provable AI inference

> Original title: Progress on provable inference for large language models

- Source: Anthropic Research
- Published: 2026-11-02T05:00:00.000Z
- Bad Influence: https://badinfluence.news/items/ebf0eqqi3vxypo110p9h9m1b2
- Original: https://www.anthropic.com/research/provable-inference

## Summary

Anthropic has released FlashProve, an open-source prototype that generates cryptographic proofs showing a response came from a committed AI model. On a single H100 GPU, the system proves Llama 4 Maverick at about 110 tokens per second, with proofs of about 19 MB per 1,024-token request. The release includes code, a technical report, and Lean formal proofs of its security properties, and the proofs are limited to 1,024-token contexts.

## Why it matters

The post explains how cryptographic proofs can verify which model served a response, with concrete speed and proof-size measurements that help readers judge its practicality.