Lagrange Labs proves Google’s Gemma3 model with zero-knowledge cryptography, showing verifiable AI can keep pace with frontier innovation.
Lagrange Labs announced a breakthrough at Google’s Zero-Knowledge event in Singapore, revealing that its platform, DeepProve, has successfully generated the world’s first cryptographic proof of Google’s Gemma3 large language model inference.
This achievement demonstrates that cutting-edge AI architectures can combine with cryptographic guarantees of correctness, without compromising performance or innovation speed. It represents a significant inflection point for both artificial intelligence and zero-knowledge (ZK) technologies, highlighting that verifiable AI can evolve in real time with the most advanced research.
“Gemma3 represents a clear departure from older GPT-style models with smarter attention mechanisms and novel optimisations. By proving Gemma3, we’re demonstrating that verifiable AI isn’t a retrofit for yesterday’s models. It’s living infrastructure that evolves with the frontier of AI research,” said Ismael Hishon-Rezaizadeh, CEO and co-founder of Lagrange Labs.
Technical Innovation
Proving the Gemma3 architecture required breakthroughs in DeepProve’s cryptographic system. Adjustments were made to handle Group Query Attention, where Key and Value tensors are shared across Query heads, as well as Local and Global Attention layers that demand flexible proof structures.
DeepProve also adapted to Rotary Positional Encoding (RoPE) for more efficient scaling, and integrated RMSNorm layers into its proof circuits while maintaining verification speed.
Regulatory and Industry Impact
The ability to prove advanced AI models cryptographically is expected to transform compliance-heavy industries. Healthcare organisations could verify diagnostic models, defence firms can validate mission-critical AI systems, and financial institutions can provide cryptographic receipts for algorithmic decisions.
Organisations no longer need to choose between innovation and verifiability. DeepProve on Gemma3 proves that compliance keeps pace with model upgrades,” added Hishon-Rezaizadeh.
Scaling for Production
DeepProve’s Gemma3 integration builds on its production record of more than 11 million zero-knowledge proofs generated and 3 million AI inferences verified. The system maintains a 158x performance advantage over competing zkML solutions, while supporting the most advanced model architectures.
“With DeepProve on Gemma3, proving AI correctness isn’t hypothetical anymore. It’s cryptographically guaranteed,” Hishon-Rezaizadeh said.
This milestone strengthens the case for verifiable AI as an essential foundation for the future of secure, compliant, and scalable artificial intelligence.
Stay updated on the crypto space by following our socials
