

Protenix-v2
#5 v Věda a bio AIByteDance · v2 · od 2026-04-08 · 4× · naposledy 13. 11. 2026
56
Momentum
Protenix-v2 is a biomolecular structure prediction model (about 464M parameters) from ByteDance's Seed Team, released on April 8, 2026. According to the official repository, it improves antibody-antigen complex prediction over Protenix-v1 and updates ligand-related plausibility. The licensing is contradictory: the GitHub README calls the weights proprietary, while third-party sources state Apache 2.0.
Vývoj momenta
18.08.16.11.
Vlastnosti
| Deployment Model | Local via pip or Docker (GPU) plus hosted web server |
| Use Case Scope | Structure prediction of protein, DNA, RNA and ligand complexes; antibody-antigen prediction; ligand plausibility |
| Integrations | Web server (protenix-server.com), PyPI package, Docker image; part of Protenix family incl. PXDesign |
| License | Conflicting: GitHub README says v2 weights are proprietary, not under any open-source license; third parties state Apache 2.0 |
| Platform | Python >=3.11, PyTorch; pip install; NVIDIA GPU with CUDA required (linux x86_64) |
| Price | Protenix server free for all uses incl. commercial (per provider; v2 weights only via server) |
| Release Date | April 8, 2026 (Protenix-v2); PyPI protenix 2.0.0 on April 7, 2026 |