Open-Source and Open-Weight LLMs
Open-source and open-weight language models are often described interchangeably. Downloadable parameters alone do not establish that an entire AI system meets an open-source definition. Inspect weights, code, data information, and licensing separately.
Distinguish the terms
Open weights usually means model parameters are available, with permissions governed by their license. OSI's Open Source AI Definition 1.0 requires freedoms to use, study, modify, and share, plus a preferred form for modification including data information, code, and parameters. It does not simply require publishing all raw training data.
This entry uses OSI's terminology; vendors, communities, and other standards may use different definitions.
Suggested review
| Item | Question |
|---|---|
| Weights | Can you obtain a fixed revision? |
| Code | What inference and training code is provided? |
| Data information | Are provenance and processing documented? |
| Licenses | What terms apply separately to each component? |
| Reproduction | Can you reproduce results on your own task? |
Check the original files for the specific release rather than inferring permissions from a model name or download button.
Possible uses
Self-hosting, offline inference, and fine-tuning depend on technical requirements and version-specific permissions. Our suggested selection record stores the revision, license links, and evaluation configuration so upgrades remain traceable.
Sources
- OSI: Open Source AI Definition 1.0 — definition and required components.
- OSI: Open Weights — distinction from open-source systems.