Training / Standard term
Open weights vs open source
Open weights provides downloadable learned parameters under a license; open-source AI additionally makes the system's preferred form for modification available under an open definition.
A weight release may allow local inference and fine-tuning without disclosing training code, data information, or a reproducible pipeline. The Open Source Initiative's Open Source AI Definition requires freedoms to use, study, modify, and share plus access to specified code, parameters, and sufficiently detailed data information. Other organizations use broader or narrower definitions, so the license and released artifacts remain decisive.
Builder example
Self-hosting, redistribution, commercial use, fine-tuned derivatives, acceptable-use terms, attribution, and patent clauses vary by release. Download access does not imply permission to do anything, and an open license does not by itself establish security, privacy, or training-data quality.
Common confusion: Open is contested vocabulary, but that does not make every definition equally precise. State whether you mean weights, code, data information, license freedoms, or reproducibility.