NEAR plans to utilize this verification agent to formally prove its own protocol and core smart contracts, as well as release a tool for other developers to verify their own software. The achievement follows a previous first-place finish on the putnambench leaderboard and is part of a broader investment to make software verification affordable and scalable for a wider range of users.
Tuesday, October 20, 2026
Story
NEAR AI Tops Lean Eval v1 After 4 Weeks of Competition
AI18 days ago
An AI agent developed by the protocol's research team outperformed all other participants on a leaderboard managed by Lean FRO. The system, which is almost entirely powered by the open-weight DeepSeek V4.1 Flash model, claimed the No. 1 spot on the Lean Eval v1 benchmark for difficult formalization problems just four weeks after joining the competition. This architecture allows the division to operate without third-party model dependencies and significantly reduces the cost of formal verification.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.