Open Source vs. Open Weights

Open source software traditionally means the full source code, training data, training procedures, and weights are all freely available, modifiable, and redistributable. Applied strictly, almost no AI model meets this definition, sharing training data involves enormous legal complexity around copyright and data licensing.

Open weights means the trained model parameters are publicly downloadable, but the training data and full training procedures may not be disclosed. Meta's Llama series, Mistral, and Qwen are open weights. You can download, run, modify, and fine-tune them, but you can't recreate the exact training run.

The distinction matters for legal, business, and reproducibility reasons. An "open weights" model is still commercially valuable (you can self-host it without API costs) but it's not fully open in the traditional sense.