To develop a generative model, you will need to follow these general steps:

Collect and preprocess the data: Collect a large dataset of the type of data you want to generate (e.g. text, music, images). Preprocess the data to make it suitable for training a generative model.

Choose a model architecture: There are various architectures for generative models such as Generative Adversarial Networks (GANs), Variational Autoencoders (VAEs), and Transformer-based models (such as GPT-3). Choose the one that is most appropriate for the type of data you want to generate.

Train the model: Use the preprocessed data to train the generative model. This typically involves using an optimization algorithm such as stochastic gradient descent (SGD) to adjust the model's parameters so that it can generate data that is similar to the training data.

Evaluate the model: Evaluate the performance of the model by comparing the generated data to the training data. Metrics such as the Inception Score (IS) and Fréchet Inception Distance (FID) can be used to evaluate the quality of the generated images.

Fine-tune and deploy the model: Once the model is trained, you can fine-tune it on a smaller dataset to improve its performance. Finally, deploy the model in a production environment where it can be used to generate new data.

Keep in mind that developing generative models such as GPT-3 can be complex and computationally intensive. It requires a significant amount of data, computational resources and domain expertise.

It's also worth noting that GPT-3 is a pre-trained model developed by OpenAI and it's not open source for general use. However, you could use similar architectures and train your own model using a dataset of your choice.