Skip to main content

Model basics

What are Microsoft DeepDev models?

Microsoft DeepDev models are a suite of state-of-the-art language models, encoders, translation models, and more. They are trained on vast collections of open source code and natural language documents. These models have been built from the ground up to understand source code tasks, and can boost progress in developing machine learning solutions for software engineering tasks.

In addition to pre-trained, task agnostic models ready to be fine-tuned for any desired task, Microsoft DeepDev contains a library of task-specific models like code completion, unit test generation, function generation, code summarization, merge conflict resolution, and more.

Further, our team of engineers and scientists are working to continuously expand the repertoire of Microsoft DeepDev model tasks, architectures, and unique performance metrics.

Model information

Every model is identified by its name, version, and owner name. As you query the models, you will need to provide this information. For more information on how to discover the models, see Discover Models