🚀

Featured: EGIP Release

April 24, 2025 • Radical AI

EGIP: Another step forward in ML-first materials

Announcing Efficient Geometric Interatomic Potential (EGIP), our first-generation interatomic potential that achieves state-of-the-art accuracy on thermal conductivity prediction while being 3.5x faster than leading models. EGIP demonstrates that direct force prediction models can excel at materials property prediction when built correctly.

🏆 State-of-the-art thermal conductivity ⚡ 3.5x speed improvement 🔬 Direct force prediction
Read Full Article

Featured: TorchSim Release

April 4, 2025 • Radical AI

Introducing TorchSim: A next-generation atomistic simulation engine

Announcing TorchSim, a next-generation PyTorch-native atomistic simulation engine for the MLIP era. TorchSim revolutionizes atomistic simulation to unlock 100x speedup over ASE and 100,000,000x speedup over DFT, with full GPU utilization and parallel system evolution.

🚀 100x speedup over ASE 🔥 PyTorch-native 🌐 Open source
Read Full Article