metaai·lightalo unofficial · independent
Universe / AI for Science / ESM / ESMFold
AI for Science · 2019-2022

ESM / ESMFold

ESM / ESMFold (Evolutionary Scale Modeling)

617 million protein structures in two weeks

open source archived ESM-2 8M to 15B; ESMFold 690M + 3B ESM-2 backbone params

Latest: ESM-2/ESMFold (2022); Meta repo archived August 2024; ESM3 is EvolutionaryScale's, not Meta's

Meta's protein language models: ESM-2 (up to 15B parameters, 2022) learned protein structure from sequences alone, and ESMFold predicted atomic-level 3D structure ~60x faster than AlphaFold2 without multiple sequence alignments (Science, 2023). The core team left Meta in 2023 to found EvolutionaryScale, whose ESM3 (2024) continues the family — outside Meta.

Why it matters

Showed protein language models trained on sequences alone learn structure: ESMFold predicted 3D structure about 60x faster than AlphaFold2 without multiple sequence alignments, enabling the ESM Metagenomic Atlas. The core team left to found EvolutionaryScale in 2023, and Meta archived the repository in August 2024.

Facts

Try it yourself

Sources

More in AI for Science

Open Catalyst ProjectBrain & AI: Decoding Perception from MEG/EEGOpenDAC / ODAC25Open Materials 2024fairchemOpen Molecules 2025

Read the AI for Science story on the sky →

✦ Open on the map Explore AI for Science Quiz me