“Open weights” means the model’s trained parameters — the numbers that actually encode what it learned — are published for anyone to download and use. You can run it on your own hardware, inspect it, and fine-tune it for your task, without a per-token dependency on a closed provider.
It’s distinct from “open source.” The Open Source Initiative’s Open Source AI Definition requires the weights plus the complete code used to train and run the model and enough information about the training data for a skilled person to build a comparable system. Many “open” models release only the weights, sometimes under licenses that restrict how they can be used. That distinction matters for reproducibility and licensing, but for most builders the practical win is the same: open weights let you self-host, keep data in-house, and customize — the freedoms that make the open ecosystem a real alternative to proprietary frontier APIs.
Sources
- Open Source Initiative: The Open Source AI Definition 1.0 — Requires data information, the complete training and inference code, and the model parameters, so released weights alone do not meet the definition; checked October 25, 2026.
Go deeper
- Should you use open source or proprietary LLMs? AI, decoded · Open vs. Proprietary AI Models: When to Use Which
- Is an open-weight AI model the same as an open-source one? AI, decoded · Open-Weight vs. Open-Source AI Models
- Can switching to an open model actually cut your AI costs? AI, decoded · When an Open Model Beats a Frontier API on Cost
