ByteDance's open-source Protenix overtakes AlphaFold3 across FoldBench benchmarks
RichardsonDx · x · 2026-10-25
Google's AlphaFold3 predicts protein structures and drug binding; ByteDance rebuilt it as the open-source Protenix, which now beats AlphaFold3 on matched benchmarks. FoldBench-Corrected results: monomer lDDT 88.6 (PX-v1, 4 seeds) vs 88.0 for AF3; antibody-antigen docking success 55.6% vs 48.8%; protein-ligand success 63.5% vs 62.6%, also well ahead of Boltz-1, Chai-1 and HelixFold3. Since AF3 was closed, Protenix lets any lab or startup run and build on state-of-the-art structure prediction themselves.
Related event: ByteDance's open-source Protenix outperforms AlphaFold3(2 posts)→
More from Models
- AI Writing Is Getting Harder to Read — Research Says It's Trainable — lmoroney · 2026-10-25
- ChatGPT web quietly adds a usage limit popup showing Pro plan quota — menhguin · 2026-10-25
- 4B model fine-tuned 27 min on one A100 lifts letter-exam accuracy to 74.5% — MaziyarPanahi · 2026-10-25
- Even at temperature 0, 45 of 1,273 LLM answers changed between runs — MaziyarPanahi · 2026-10-25
- Claude reportedly went off-script in UK AISI hacking test, 9 of 43 runs — faraz0x · 2026-10-25
- Frontier LLMs can build their own scoring harness to hill-climb any goal — Ryoiki-Tokuiten · 2026-10-25