From the source
MiniMax's research team used an internal-only research model, MIM, to run automated cryptanalysis across modern cryptographic schemes, finding previously unknown structural flaws and turning theoretical attacks into practical breaks.
- The work spans three areas: NIST standardization candidates, peer-reviewed constructions and competition winners, and classical hard mathematical problems underlying public-key cryptography.
- On CryptanalysisBench Tier 2, which holds 135 publicly reviewed schemes with no known practical attacks, MIM produced the first practical collision attack on Waterfall-256, finding collisions across all output lengths in 0.045 seconds on a single core.
- It also broke full-round collision resistance in Blender-256, finding collisions in 0.36 to 2.3 seconds, and recovered secrets for the post-quantum candidates LOCKER and ROLLO-I-128.
- MIM found more than a dozen unreported implementation bugs in Tier 2 reference code, including a verification bug in DRS that allowed signature forgery with zero signing queries.
- On classical hard problems, MiniMax says MIM solved dozens of previously unsolved high-order instances in public challenges, with measured speedups of hundreds to tens of thousands of times over prior state-of-the-art solvers.





