---
title: Open Models Now Beat AlphaFold3 at Predicting How Antibodies Bind Their Targets
description: ESMFold2, Protenix-v2 and OpenDDE all cleared AlphaFold3's published FoldBench score on antibody-antigen structures in independent testing by Tamarind Bio.
author: Darie Nani (Editor-in-Chief)
updated: 2026-08-11T05:21:19.308Z
canonical: https://www.sovereignmagazine.com/article/open-models-beat-alphafold3-antibody-antigen
image: https://cdn.nanimediahouse.com/open-models-alphafold3-antibody-137813.webp
categories: Science & Tech
content_type: News
region: Global
publication: Sovereign Magazine
access: members
schema_type: Article
---

Three [openly released](https://www.sovereignmagazine.com/article/meta-muse-glimmer-open-30b-apache-2) structure prediction models now outperform Google DeepMind's AlphaFold3 at predicting how an antibody binds its target, according to [independent testing](https://www.sovereignmagazine.com/article/irregular-frontier-ai-security-lab-openai-anthropic-meta) by Tamarind Bio, a computational biology platform that ran all three on the same inputs and setup.

Tamarind Bio writes that over the past year many groups, mostly companies, have claimed a new model matched or outperformed AlphaFold3 on some structure prediction task, and that a substantive number of those claims did not survive independent scrutiny. Its author says he was initially skeptical of the results from the ESMFold2, Protenix-v2 and OpenDDE teams, whose releases came within a few weeks of each other.

ESMFold2 came from Biohub on 27 May 2026, in a release that also included ESMC and ESM Atlas and that Biohub says predicts protein and biomolecular complex structures with state of the art accuracy and speed. Protenix-v2 is ByteDance's open-source model, released on 8 April 2026 with what ByteDance describes as clear gains on antibody-antigen structure prediction, following Protenix-v1 on 5 February 2026. OpenDDE is the open-source version of AuraIDE, the proprietary foundation model built by Aureka Biotechnologies.

---

*This article is only available to registered readers. Visit the article URL to read the full piece.*