PoET-2
#39 v Věda a bio AIOpenprotein Ai · v2 · 9× · naposledy 30. 6. 2026
1
Momentum
PoET-2 is a multimodal protein language model by OpenProtein.AI (NE47 Bio) that learns generative distributions over protein sequences through in-context Learning from sequence and optional structure data. With only 182 million parameters, it achieves performance comparable to trillion-parameter models according to the provider and substantially outperforms larger Foundation Models at a fraction of the computational cost. The model is offered via the OpenProtein.AI web platform and APIs for tasks such as Zero-Shot variant effect prediction and sequence property modeling; code and model weights are available on GitHub/Zenodo for non-commercial purposes.
Vývoj momenta
19.05.17.08.
Vlastnosti
| Deployment Model | Cloud platform (web app/API); inference code available for self-hosted Linux/NVIDIA GPU environment via GitHub |
| Use Case Scope | Protein engineering: zero-shot variant effect prediction, sequence generation, fitness/stability/binding prediction |
| Integrations | OpenProtein.AI platform incl. AlphaFold2, ESM2 and Clustal Omega for design & optimization |
| License | PoET Non-Commercial Artifacts License (model weights); commercial license available on request from NE47 Bio |
| Platform | OpenProtein.AI web app and REST/Python API |
| Price | Free for academic use; commercial use via OpenProtein.AI platform/API (pricing not publicly listed) |
| Release Date | February 11, 2025 (announcement) |