Chai-1 (weights used by chai_lab 0.6.1)
Release in the Chai-1 family · version chai_lab 0.6.1 (models_v2 components)
Maintained by Chai Discovery186
The Chai-1 weights used by the current chai_lab package (version 0.6.1, published March 18, 2025). They are distributed as six exported PyTorch module files (feature embedding, bond projection, token embedder, trunk, diffusion module, and confidence head). The package fetches them from Chai's asset server, and the same six files are in the Hugging Face repository, last updated February 18, 2025. The sources reviewed do not say which earlier package versions used these same files.25347
- Model hub: Chai-1 weights (Hugging Face) (external site: huggingface.co)
- Repository: GitHub repository (chai-lab) (external site: github.com)
- License: LICENSE (Apache 2.0) (external site: github.com)
- Paper: Chai-1 technical report (bioRxiv, doi:10.1101/2024.10.10.615955) (external site: doi.org)
Availability and license
Overall availability
The chai_lab package downloads the weights automatically from Chai's asset server, and the same component files are in an ungated Hugging Face repository. The current LICENSE and README do not reference an acceptable use policy; the original 2024 community license did.37189
Availability is separate from permission: read the license before using or redistributing.
Current terms: the repository LICENSE is Apache License 2.0 (copyright 2024 Chai Discovery), the README states that code and model weights are both under Apache 2.0 for academic and commercial use including drug discovery, and the Hugging Face repository is tagged apache-2.0. Earlier terms: at the September 2024 release, LICENSE.md was the Chai Discovery Community License Agreement from Chai Discovery, Inc., which granted rights to the code, weights, and outputs only for non-commercial purposes, excluded commercial entities, and incorporated an acceptable use policy. That file was replaced with the Apache 2.0 text in a commit dated November 27, 2024, and package version 0.4.2 was released the same day. This release record covers the weights fetched by version 0.6.1, which postdate the change.81791110125
Component reuse rights
- weights
- Reviewed qualifying license recorded — check scope and conditions
- code
- Reviewed qualifying license recorded — check scope and conditions
- data
- Unknown — no complete fact-level rights review
- documentation
- Unknown — no complete fact-level rights review
No complete system-rights review is recorded for this release.
Model-disclosure tier
The model parameters for this release can be downloaded by the public. License terms may still restrict use, redistribution, or commercial use.
Public materials checklist
| Item | Status | Notes and evidence |
|---|---|---|
| WeightsCan the general public download the model parameters for this release? | Public | Six exported component files, downloadable without gating from Hugging Face and fetched automatically by the package.73 |
| Inference codeIs code for running the model published? | Public | The chai_lab package provides the chai-lab fold command and Python entry points.15 |
| Training codeIs the code used to train the model published? | Unknown | The README covers inference only; no training code was found in the sources reviewed.1 |
| Training-data informationDoes the information cover provenance, scope, acquisition, selection, labeling, processing, and where data or alternatives can be obtained? Access alone does not establish completeness. | Unknown | The technical report was not reviewed for this record. |
| Training-data accessCan the training data be obtained? This is independent of information completeness and reuse rights; original unshareable data need not be downloadable. | Unknown | Not assessed. |
| Complete training pipelineIs the complete base-training and preprocessing pipeline published, including configuration? Fine-tuning code or an inference SDK alone is insufficient. | Unknown | Not assessed. |
| Legacy data assessment (v0.1)Historical assessment combining download access and disclosure. Preserved for traceability; excluded from the v0.2 tier calculation. See the new separate assessments above. | Unknown | Not assessed. |
| Training recipeAre the training configuration and procedure documented in enough detail to follow? | Unknown | The technical report was not reviewed for this record. |
| Evaluation materialsPublic = evaluation code or prompts that let others re-run the evaluations are published. Partial = results only. | Unknown | The README shows a benchmark chart and links the technical report; no evaluation scripts were identified in the README. |
What it is useful for
Folding complexes of proteins, small molecules, DNA, RNA, and glycosylations from FASTA input, optionally with MSAs, templates, inter-chain contact or covalent-bond restraints, and custom embeddings.1
Run and use notes
- The README says the package requires Linux, Python 3.10 or later, and a CUDA GPU with bfloat16 (bf16) support. It recommends A100 80GB, H100 80GB, or L40S 48GB GPUs, with A10 and A30 usable for smaller complexes; it states no further memory assumptions, such as complex size.1
- By default the model generates five samples without MSAs or templates; the README recommends MSAs and documents an option to query the public ColabFold MMseqs2 server.1
Organization context
Other releases in the Chai-1 family
No other releases in this family have been assessed.
What this catalog does not know
- Release date: not documented in the sources reviewed.
- Training code: unknown.
- Training-data information: unknown.
- Training-data access: unknown.
- Complete training pipeline: unknown.
- Legacy data assessment (v0.1): unknown.
- Training recipe: unknown.
- Evaluation materials: unknown.
Have a primary source? How to report a correction.
U.S. eligibility
Eligible · basis: U.S.-governed project
Developed and published by Chai Discovery in its GitHub and Hugging Face organizations; the LICENSE and package source name Chai Discovery as copyright holder. Chai Discovery's careers page places its team in San Francisco, California, and its website terms choose California law with San Francisco courts.8261314
Sources
This listing is not an endorsement, a safety assessment, or a federal approval.