

Protenix
#6 v Věda a bio AIByteDance · od Januar 2025 (erste Version); Protenix-v2 am 8. April 2026 / January 2025 (first release); Protenix-v2 on April 8, 2026 · 4× · naposledy 13. 11. 2026
54
Momentum
Protenix is an open-source model for predicting three-dimensional biomolecular structures (proteins, nucleic acids, ligands, ions), developed by ByteDance's Seed team / ByteDance Research as a reproduction of AlphaFold 3. The current version, Protenix-v2 (464M parameters), was released on April 8, 2026; according to the GitHub README its weights are proprietary, while code and weights of the other versions are under Apache 2.0.
Vývoj momenta
18.08.16.11.
Vlastnosti
| Deployment Model | Local via pip/Docker with NVIDIA GPU, or via web server; v2 weights only via web server |
| Use Case Scope | 3D structure prediction of proteins, DNA/RNA, ligands, ions; antibody-antigen complexes; VHH design (v2) |
| License | Apache 2.0 (code and parameters); Protenix-v2 weights proprietary per README, not open source |
| Platform | Python/PyTorch codebase (pip, GitHub) plus web server protenix-server.com |
| Price | Protenix web server free, incl. commercial use |
| Release Date | Protenix-v1: Feb 5, 2026; Protenix-v2: Apr 8, 2026 |