# HannesStark/boltzgen

BoltzGen: Toward Universal Binder Design

Repository: https://github.com/HannesStark/boltzgen
Canonical: https://ross.abutalabs.com/products/boltzgen
Homepage: https://boltz.bio/boltzgen
Language: Jupyter Notebook
License: MIT
License Family: permissive
Topics: antibody-design, antibody-engineering, denovo-design, drug-design, drug-discovery, peptide-design, protein-engineering, structural-biology, miniprotein-design, nanobody-design
Last push: 2026-05-28T19:33:10+00:00

## Health v2 (maintenance only)
Score: 71/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 84, release rhythm 81, longevity 22
- inputs: {"age_days": 311, "days_push": 97, "days_rel": 124, "gap_med": 6, "n_releases_24m": 10}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1042, forks 253 (observed 2026-08-28T04:03:20.518336+00:00)

## What it is
BoltzGen is an open-source all-atom generative diffusion model for designing protein and peptide binders against arbitrary biomolecular targets (proteins, nucleic acids, small molecules). It unifies binder design and structure prediction in a single model, controlled via a YAML design specification language, and is installable via pip or Docker.

## Use cases
- design protein binders against a target protein
- generate nanobodies or antibodies de novo
- design peptides that bind nucleic acids or small molecules
- predict structures of designed proteins
- run binder design pipelines on GPU with Docker
- rank candidate protein designs for drug discovery

## When to choose
- you need de novo binder design for proteins, nucleic acids, or small-molecule targets
- you want an MIT-licensed, freely usable alternative to closed protein-design models
- you need unified design plus structure prediction in one model
- you want a flexible YAML-driven design specification workflow

## When to avoid
- you lack GPU compute or cannot download ~6GB of model weights
- you need small-molecule ADMET or affinity-only prediction rather than binder generation
- you want a hosted API without local infrastructure
- your task is unrelated to biomolecular design or structure prediction

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-inference, sdk, cli
- domain: bioinformatics, healthcare, artificial-intelligence, deep-learning, gpu-computing
- platform: python, windows, cli
- tags: protein-design, binder-design, antibody-design, nanobody-design, peptide-design, drug-discovery, generative-diffusion-model, structure-prediction, computational-biology, structural-biology, de novo protein binder generation, design specification yaml, docker, gpu, linux, macos

## Member repositories
- HannesStark/boltzgen (main) score 71

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:20.518336+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T07:02:33.697760+00:00, confidence not recorded.
  - readme: https://github.com/HannesStark/boltzgen (fetched 2026-08-28T04:03:20.518336+00:00, sha e174e6910b58)
  - homepage: https://boltz.bio/boltzgen (fetched 2026-08-29T13:03:54.761202+00:00, sha 2863bd71c8c8)
  - site_page: https://boltz.bio/pricing (fetched 2026-08-29T13:03:54.763795+00:00, sha 122e226536b8)
- Data as of 2026-08-30T08:39:29.467469+00:00.