Defined scope
- Model weights
- Local inference
- Licensing
- Quantization
- Adaptation
CONCEPT
Open-Weight Model
An open-weight model is a machine-learning model whose learned parameters are distributed for reuse under stated terms, without necessarily providing the training data, training code or freedoms required by open-source definitions.
Open-weight models enable local inference and adaptation but must be distinguished from fully open-source AI systems.
active
research
EDITORIAL FRAME
What this entry establishes.
A concise view of its scope, position, limitations and supporting sources.
Editorial position
- Open weights can improve autonomy and inspectability without providing full training reproducibility.
- Model licenses, data information, code availability and use restrictions must be evaluated separately.
Explicit limits
- A claim that downloadable weights automatically make an AI system open source
- A hosted API with no distributable model parameters
Tags and disciplines
Open WeightsLocal AIModel LicensingQuantizationArtificial IntelligenceOpen SourceProgrammingDigital Independence
References behind this page
- The Open Source AI Definition 1.0 Open Source Initiative
Definition
An open-weight model makes learned parameters available for download and inference. Availability may permit local deployment, evaluation, quantization or fine-tuning, but the exact freedoms depend on the license and supplied components.
Open source distinction
The Open Source AI Definition requires freedoms to use, study, modify and share, together with the preferred form for modification. A weight release may omit training data information or training code and therefore remain open-weight rather than fully open source.
Applications
Open-weight models support private local assistants, offline tools, edge deployments, reproducible experiments and specialized creative systems.
Electronic Artefacts position
Local and open-weight models are relevant where archives, unreleased audio, project documents or private graphs should remain under operator control. Deployment choices must still account for provenance, security, performance and licensing.
Limitations
Running weights locally does not remove model bias, hallucination, data rights concerns or maintenance cost. Hardware requirements, context limits and update procedures remain operational constraints.
References
See the OSI Open Source AI Definition, llama.cpp, Open Source and Provenance.
DOCUMENTED RELATIONSHIPS
Connected work and ideas.
Each link names the relationship between two entries and why it matters.
Local and Open Source AI Systems
Local and Open Source AI Systems documents the distinction between open-weight and open-source models.
WebNN and Local AI in the Browser
WebNN and Local AI in the Browser documents Open-Weight Model as one of its declared subjects.
Knowledge Hub Third Wave
Open-Weight Model is an explicit member of the Knowledge Hub Third Wave collection.
Record details Metadata, sharing and citation
Open-Weight Model
Cite this page
Open-Weight Model. 1.0.0. Electronic Artefacts, 2026-06-24. https://electronicartefacts.com/knowledge/concepts/open-weight-model/
Nearby relationships
3 public links connect this page to nearby projects, concepts and references.