OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Open AF3

Open AF3 is an open-source, MIT-licensed Python and PyTorch implementation of the AlphaFold 3 research approach for predicting biomolecular interactions. It includes model components for pair and single representations, pairformer processing and diffusion over atomic coordinates, along with examples for package installation, tensor construction and a model forward pass.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Open AF3 — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Open AF3 when a self-hosted Python and PyTorch implementation is more useful than a hosted interface, especially for architecture study, educational work or exploratory biomolecular modelling. The repository’s documented examples are appropriate for inspecting tensor shapes and model calls; biological prediction workflows require additional implementation and input handling beyond those examples.

Who it is for

  • Biomolecular structure researchers needing a local implementation to study or adapt.
  • Python and PyTorch developers investigating AlphaFold 3-style model components.
  • Educators and learners examining diffusion-based biomolecular modelling workflows.

Documented uses

  • Install the Python package in a local environment.
  • Construct the documented PyTorch input representations.
  • Instantiate the AlphaFold3 model and inspect a forward-pass output.
  • Study or adapt architecture components related to pair representations, pairformer processing and coordinate diffusion.
Before you choose

Requirements & limits

The repository is public, open source and MIT licensed. Its README documents installation with `pip install alphafold3`.

  • The implementation is Python-based and uses PyTorch.
  • It is self-hosted rather than presented as a hosted prediction service.
  • The repository documents model examples but also lists unfinished input, template, MSA, residual, recycling and training-pipeline work.
  • The documented examples use random tensors, so they demonstrate model interfaces rather than a complete biological prediction pipeline.
  • The repository describes confidence-related outputs and notes the risk of plausible-looking hallucinated structures.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Open AF3 and inspect a model forward pass

Set up the Python package and reproduce the repository’s documented tensor-based model construction example.

Before you start

  • Python environment with pip
  • PyTorch
  • A local environment capable of installing the alphafold3 package

Steps

  1. Install the package with the command below.

    Documented command
    pip install alphafold3

  2. Import torch and create pair and single representation tensors using the documented shapes: (1, 5, 5, 64) and (1, 5, 64).

  3. Instantiate AlphaFold3 with the documented example parameters, including dim=64, seq_len=5, heads=8, pair_former_depth=48, num_diffusion_steps=1000 and diffusion_depth=30.

  4. Pass the tensors to the model and inspect the returned output shape.

Expected outcome

A locally constructed AlphaFold3 model receives the example PyTorch representations and returns an output tensor as shown in the repository’s forward-pass example.

Guide limitations

  • The example uses random tensors rather than biological input data.
  • The repository lists unfinished architecture and training-pipeline work, including input embedding, templates, MSA embedding, residual handling and recycling.
  • This source-based recipe does not establish prediction quality or biological validity.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
4 October 2026
Original website title
kyegomez/Open-AF3 · Official homepage title · retained in its source language
How we review sourcesSuggest a correction