# microsoft/mattergen

Official implementation of MatterGen -- a generative model for inorganic materials design across the periodic table that can be fine-tuned to steer the generation towards a wide range of property constraints.

Repository: https://github.com/microsoft/mattergen
Canonical: https://ross.abutalabs.com/products/mattergen
Homepage: https://www.nature.com/articles/s41586-025-08628-5
Language: Python
License: MIT
License Family: permissive
Topics: generative-ai, materials-design, materials-science
Last push: 2026-08-22T01:29:31+00:00

## Health v2 (maintenance only)
Score: 67/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 98, release rhythm 40, longevity 46
- inputs: {"age_days": 644, "days_push": 12, "days_rel": 406, "gap_med": 0, "n_releases_24m": 4}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1801, forks 344 (observed 2026-08-28T04:05:38.022759+00:00)

## What it is
MatterGen is Microsoft's official implementation of a generative diffusion model for designing inorganic crystalline materials across the periodic table. It ships pre-trained checkpoints and fine-tuned variants that steer generation toward property constraints such as band gap, magnetic density, bulk modulus, chemical system, and space group.

## Use cases
- generate novel inorganic crystal structures with a target band gap
- design materials with prescribed bulk modulus or magnetic properties
- fine-tune a generative materials model on custom property constraints
- sample candidate crystal structures for a given chemical system or space group
- evaluate generative models for materials discovery
- train a materials generative model from scratch on MP-20 data

## When to choose
- you need property-conditioned generation of inorganic crystal structures
- you want a research-grade, peer-reviewed model with pre-trained checkpoints
- you have a CUDA GPU and want to fine-tune on your own materials datasets

## When to avoid
- you need organic or molecular (not crystalline inorganic) materials generation
- you have no GPU access and need fast, reliable inference (Apple Silicon support is experimental)
- you want a production application rather than a research codebase

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-training, data-science, simulation
- domain: machine-learning, chemistry, artificial-intelligence
- platform: python
- tags: generative-model, diffusion, materials-design, crystal-structure-generation, pytorch, property-conditioned-generation, research-code, materials-science, science, linux, gpu, macos

## Member repositories
- microsoft/mattergen (main) score 67

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:38.022759+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 2026-08-30T03:22:13.547412+00:00, confidence not recorded.
  - readme: https://github.com/microsoft/mattergen (fetched 2026-08-28T04:05:38.022759+00:00, sha ca0f9d773594)
  - homepage: https://www.nature.com/articles/s41586-025-08628-5 (fetched 2026-08-29T11:01:22.721710+00:00, sha 693b0aaf406f)
  - site_page: https://www.nature.com/openresearch/about-open-access/information-for-institutions (fetched 2026-08-29T11:01:22.738931+00:00, sha 9631b227fce6)
  - registry_pypi: https://pypi.org/pypi/mattergen/json (fetched 2026-08-29T11:01:22.740659+00:00, sha 34abc45ba481)
  - site_page: https://www.nature.com/npg_/company_info/index.html (fetched 2026-08-29T11:01:22.735630+00:00, sha eb023da6fb7d)
  - site_page: https://www.nature.com/npg_/press_room/press_releases.html (fetched 2026-08-29T11:01:22.737320+00:00, sha 6f7e0f4fdb50)
- Data as of 2026-08-30T08:39:29.467469+00:00.