The Ramanujan Machine project has announced the Ramanujan Challenge, a new evaluation framework designed to test the mathematical reasoning capabilities of artificial intelligence. Launching on July 1, 2026, with a submission deadline of August 1, 2026, the initiative aims to determine whether current AI systems can derive and verify research-level formulas for fundamental mathematical constants, including pi, e, Catalan’s constant, and special values of the Riemann zeta function. The challenge addresses a persistent gap in AI benchmarking by offering problems that are numerically verifiable to arbitrary precision yet require non-trivial mathematical insight. Participants will tackle two categories of problems: formulas with proofs held in cryptographic escrow by the organizers, and unproven mathematical conjectures. The initiative is structured to prevent verification bottlenecks by prioritizing reproducible, machine-readable submissions. Entries will be evaluated in descending order of priority, beginning with formal proofs developed in interactive theorem provers such as Lean, Coq, or Isabelle. Second in priority are computer algebra system derivations using established symbolic libraries like Mathematica, SageMath, or SymPy. Human-readable proofs will be considered only when computational submissions are insufficient. Strict technical guidelines govern participation. Submissions must include fully documented, reproducible source code capable of verifying the mathematical steps. Any symbolic system or library utilized must have been publicly released prior to the July 1 launch date, ensuring the contest evaluates algorithmic reasoning rather than newly generated mathematical software. AI-generated code may be used as implementation scripts but cannot function as an unverified computational oracle. The organizers explicitly prohibit reliance on private APIs, hidden remote services, or unverifiable computations. All code must be accompanied by a detailed mathematical derivation in PDF or TeX format. To maintain academic integrity, participants are encouraged to engage in open discussion regarding methodology and strategy, but posting complete solutions before the August 1 deadline is strictly prohibited. Submissions can be made through the official challenge portal or via email, both of which record timestamps and participant information for transparency. Following the evaluation period, the Ramanujan Machine team will publish a comprehensive report detailing accepted solutions, partial attempts, and the specific AI systems or tools employed. Any formally verified proof submitted for an open conjecture will be recognized as a novel mathematical contribution, with authorship credited in a dedicated project publication. By establishing a standardized, verifiable framework for AI-assisted mathematics, the Ramanujan Challenge seeks to bridge the divide between computational pattern matching and rigorous proof generation. The initiative reflects a growing industry focus on moving beyond statistical approximation toward systems capable of producing logically sound, publishable mathematical research.