tarasovcode
Models1 min readMarkTechPost

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

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.

Save
Share:VKTelegramWhatsApp
Anthropic Releases FlashProve: Cryptographic Proofs That a Reply Came From Committed Weights
Illustration: Tarasov Code

In brief

  • What: Anthropic Releases FlashProve: Cryptographic Proofs That a Reply Came From Committed Weights
  • When: November 4, 2026 at 01:08
  • Source: MarkTechPost (media)
  • Topic: Models

#Anthropic#Models

Prepared by Ilya Tarasov’s Tarasov Code newsroom based on a report by MarkTechPost.

Read next

TechCrunch AI · November 4, 2026Anthropic is working on a system to prove AI models’ identityTechCrunch AI · November 3, 2026Anthropic is still grappling with the economics of computeSiliconANGLE AI · November 3, 2026Anthropic’s quarterly revenue jumps to $17B, losses grow ahead of IPOTechCrunch AI · November 3, 2026Anthropic warns that it’s very close to RSI capability, aka self-building AI