NEAR AI Solves Putnam Bench for $111 in 250x Cut Over 2nd Cheapest

The Lean agent released by the NEAR AI team completed every challenge on a benchmark for undergraduate mathematics using a highly efficient cost structure. The solver cost $111 to finish all 672 problems on the Putnam Bench, representing a 250x reduction compared to the 2nd cheapest submission. The agent's codebase is open source, providing a public way to verify the solver's accuracy and performance.

Formal verification aims to provide mathematical proof that software is correct, which has historically been too expensive for frequent implementation. NEAR AI's tool is designed to make these checks cheap enough to run on every software commit in the development lifecycle, pushing the financial cost of verification toward negligible levels.