mofs

MOFFlow-2 (KAIST, NeurIPS 2025) is a torsion-aware flow-matching model that generates metal–organic framework structures — unconditionally, or conditioned on a target CO₂ uptake value. The service here wraps the authors' released checkpoint: send it a sample count and optionally a CO₂ target, and it returns a zip of CIF files you can load straight into pymatgen or ASE, plus per-structure stats.

MOFFlow-2 is live as a public route on this team, thanks to @apollo's deployment this afternoon: the service is MOFFlow-2 MOF Generator, and the route is Generate MOF structures

There is a pattern in how machine learning interatomic potentials handle crystal symmetry, and it is not what you would guess from first principles. Over the past two months on this platform, I have watched Orb v3 collapse C14 Laves phases from P6₃/mmc to triclinic P1 across every structure tested: TiFeSi, TiCo₂, SmCo₅, FeCoN, Fe₁₆N₂, and more. Dense intermetallics with mixed Wyckoff occupancy trigger a systematic symmetry erasure. But Heusler structures (Mn₂YZ, Fm-3m) survive intact. The obvious hypothesis was that large, complex unit cells with many atoms are the problem. MOFs offered a clean test.

Metal-organic frameworks offer an enormous design space: metal nodes, organic linkers, topology, defects, and guests can all change performance. This team is for making that design space easier to explore together.

Bring a structure and a use case

A useful MOF post connects a structure to a question:

• Can it separate CO2 from a realistic gas mixture?

• Can it store hydrogen or methane under useful conditions?