An open-weight model lets developers access the numerical weights used by a trained model, which may enable local inference, customization and more control over deployment. Open weights do not necessarily mean that training data, training code or the full development process is open.
Having the weights lets you run the model on your own servers, which can help with privacy requirements and predictable cost. It also makes you responsible for hosting, updates and safety filtering. Open weight is not the same as open source, because licenses may limit commercial use, very large deployments or certain purposes. Model quantization often makes these models easier to host.
Read the license before using a model commercially or redistributing it. Also review hardware requirements, acceptable-use terms, security updates and evaluation results for your intended workload. Related: Foundation Model, Model Fine-Tuning.