---
title: AceFF-2 ML Potential Explorer
emoji: ⚛️
colorFrom: purple
colorTo: green
sdk: gradio
sdk_version: 6.30.0
app_file: app.py
short_description: Energy, forces & relaxation with the AceFF-2 ML potential
python_version: "3.12"
startup_duration_timeout: 30m
license: apache-2.0
---

# AceFF-2 — Machine Learning Potential for Small Molecules

Interactive demo for [AceFF](https://huggingface.co/papers/2601.00581)
(*"AceFF: A State-of-the-Art Machine Learning Potential for Small Molecules"*,
Acellera), a neural interatomic potential for drug-like molecules built on the
**TensorNet v2** architecture in [TorchMD-Net](https://github.com/torchmd/torchmd-net).

## What it does

- **Input** a small molecule as an **XYZ block**, **PDB**, or **SMILES**
  (SMILES gets a 3D geometry embedded with RDKit).
- Set the **net charge** (−2…+2) and pick a checkpoint:
  [AceFF-2.0](https://huggingface.co/Acellera/AceFF-2.0) or
  [AceFF-2-OMOL](https://huggingface.co/Acellera/AceFF-2-OMOL) (adds Na, K).
- **Single point**: potential energy (eV and kcal/mol) and per-atom forces
  (eV/Å), visualized as force arrows on an interactive 3Dmol.js viewer.
- **Relax**: optionally minimize the geometry with ASE's LBFGS at the AceFF-2
  potential and download the relaxed XYZ.

Supported elements: H, B, C, N, O, F, Si, P, S, Cl, Br, I (AceFF-2.0) or with
Na, K (OMOL). Small molecules only. Model and code are Apache-2.0 / MIT
licensed. Example molecules are taken from the Acellera usage docs
(caffeine, aspirin, acetate) — geometries embedded locally with RDKit.