Generative AI · Fast-moving · Intermediate
Open-Weight Model
Also known as: Open-source AI model
A model whose trained parameters are published for download, allowing self-hosting, inspection and modification under a licence.
What Open-Weight Model is
Open weights are not the same as open source: training data and code are often withheld, and licences may restrict commercial use or scale. The accurate term for most releases is open-weight.
How it works
Weights are distributed through model hubs, then run locally or on rented GPUs, usually quantised, and adapted with LoRA or full fine-tuning.
Why it matters
Open weights give data residency, cost control, offline operation and freedom from provider changes, in exchange for owning the operational burden.
Common uses
- →On-premise deployments with sensitive data
- →Cost-controlled high-volume inference
- →Research and reproducibility
- →Edge and offline products
Strengths
- ✓No per-token vendor fee
- ✓Full control and privacy
- ✓Inspectable and fine-tunable
Watch for
- ✓You own serving, scaling and security
- ✓Licence terms vary widely
- ✓Safety tuning can be removed by third parties
Continue exploring
More in this collection
Browse all AI Concepts