Deep Dive
1. Proving Gemma3 & Core Optimizations (September 2025)
Overview: This update allows Lagrange's system to verify outputs from newer, more efficient AI models like Google's Gemma3. It also makes the entire proving process faster and less resource-intensive for users.
The team adapted DeepProve to handle Gemma3's advanced architecture, making it the first zkML system to prove this model. A key optimization automatically finds and removes duplicate data tensors within a model, which cuts down on proving time and memory use. The underlying graph system was rebuilt for better reliability and to support future distributed proving networks. Additionally, a new unified layer for mathematical operations simplifies the code and improves proving speed.
What this means: This is bullish for $LA because it demonstrates the project's technical leadership in verifiable AI. The upgrades mean the network can handle more complex AI verification tasks faster and cheaper, which could attract more clients and increase demand for proof generation paid in LA.
(Source)
2. Full-Sequence GPT-2 Proofs & GPU Port (August 2025)
Overview: This upgrade significantly increased the amount of data DeepProve can verify in a single proof, achieving a 25x improvement in processing speed. It also laid the groundwork for using powerful graphics cards to accelerate computations.
The milestone was proving a full sequence of 1024 tokens for GPT-2, showcasing the system's scalability. The team refactored the core to use a newer, more efficient cryptographic library, which doubled the proving speed and reduced memory use by 10x. A new framework was introduced to smartly manage memory, making the prover work on everything from small devices to large server clusters. Work also began to port the AI inference logic from CPU to GPU.
What this means: This is bullish for $LA because it proves the network's capacity to scale efficiently. Faster and more scalable proofs make the service more attractive to developers and enterprises, potentially driving higher network usage and demand for the LA token.
(Source)
Conclusion
Lagrange's recent codebase evolution focuses squarely on scaling its zkML infrastructure to handle state-of-the-art AI models more efficiently, positioning it as a high-performance layer for verifiable computation. How will these technical advancements translate into measurable on-chain demand for the Lagrange Prover Network?