Chai
Chai DiscoveryChai-1 is Chai Discovery's multi-modal model for molecular structure prediction, giving unified predictions over proteins, small molecules, DNA, RNA, glycosylations and modified residues. It runs from a FASTA file with optional MSAs and templates, and uniquely accepts user-specified experimental restraints, inter-chain contacts and covalent bonds at several resolutions, to guide the folding of a complex.
Openness
3 high confidence- weights
- open(downloaded automatically by the chai-lab package and covered by the repository's Apache-2.0 grant)
- data
- closed(training data not released
- code
- partial(inference CLI and feature-context API
- license
- Apache-2.0(OSI)
The repository license is Apache-2.0, which on its own would say nothing about the weights, and the check that matters here is that the README extends the same grant to them in as many words. Chai-1 is therefore commercially usable weights on a closed corpus, which is the category's centre of gravity.
- https://api.github.com/repos/chaidiscovery/chai-lab recorded 2026-09-10
license spdx_id: Apache-2.0; archived: false; fork: false; no training entry point in the published surface.
- https://raw.githubusercontent.com/chaidiscovery/chai-lab/main/LICENSE recorded 2026-09-10
Apache License Version 2.0 body, Copyright 2024 Chai Discovery.
- https://raw.githubusercontent.com/chaidiscovery/chai-lab/main/README.md recorded 2026-09-10
"Chai-1 is released under an Apache 2.0 License (both code and model weights), which means it can be used for both academic and commerical purposes, including for drug discovery"; weights are downloaded automatically into the package directory; the package runs inference only.
Adoption
2 medium confidenceWeights ship through the package rather than the Hub, so there is no download count; GitHub stars are the fallback signal.
- https://api.github.com/repos/chaidiscovery/chai-lab recorded 2026-09-10
stargazers_count: 1993; archived: false; fork: false.
Capability
2 high confidenceOne attribute. Restraints are the genuinely distinctive input here and they count toward the same one - they widen what can be fed in, not what the model can be turned into, and the five sample predictions the CLI emits are samples of a prediction rather than generated designs.
- https://raw.githubusercontent.com/chaidiscovery/chai-lab/main/README.md recorded 2026-09-10
MSAs, templates, embeddings, covalent bonds and inter-chain restraints are all documented inputs; "Chai-1 uniquely offers the ability to fold complexes with user-specified restraints"; no fine-tuning or training documentation.
Verified 2026-09-10