#5959th of 83 in Open modelsFor science, health and research, science
boltz
Open biomolecular structure and binding affinity prediction models
- Stars
- 4.3k
- License
- MIT
- Last commit
- May 2026
- Last release
- Sep 2025
- Language
- Python
Missed the latest GitHub check; these facts may be older.
Overview
Boltz is a family of open biomolecular interaction models, Boltz-1 and Boltz-2, run through a `boltz predict` CLI that reads YAML inputs. Boltz-2 predicts complex structures and ligand binding affinity, reporting a binder probability and a log10(IC50) value. Code and weights are MIT licensed and install from PyPI.
Who it is for: Computational biologists and drug-discovery teams
Strengths
- Code and weights both MIT-licensed, usable commercially
- Boltz-2 outputs affinity: binder probability and log10(IC50) value
- Batch prediction from a directory of YAML files with one CLI command
- Uses NVIDIA cuEquivariance kernels on recent GPUs; Tenstorrent via a fork
Weaknesses
- Boltz-2 training code and evaluation scripts are not yet released
- CPU mode works but is significantly slower than GPU
- The --use_msa_server option sends sequences to a remote MSA server
- No Docker image or compose file provided
What it needs
- GPU optional
- Needs Python, MSA server (via --use_msa_server), NVIDIA cuEquivariance (optional)
- Models: Boltz-1, Boltz-2
Also in Open models
See all 83| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier returning typed decisions with calibrated odds | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 82 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 78 out of 100 |
| #5 | open_clipPyTorch library for training and running CLIP-style image-text models | 78 out of 100 |