Phi-4
Phi-4 is the 14B text model from which Microsoft's later Phi-4 reasoning releases were developed. Its 16K context is much smaller than the context of Phi-4-mini-instruct, despite the larger parameter count.
What distinguishes this release
The card describes a mix of synthetic material, filtered public sources and acquired academic content. It identifies English as the main training language and notes that much of the coding data uses Python and common Python libraries.Phi-4 official model card
Build the task around its input and output
Practical guidance. Use this model for a bounded text task with a clearly defined result: a short explanation, a code change or a document answer. Keep the relevant input close to the question. A larger parameter count does not make this release a substitute for the Mini model's longer input limit.
Review the result in its intended setting
Practical guidance. For code work, inspect library calls and package versions against the actual project. For document work, ask for source passages alongside the answer. Measure whether retrieval improves a task before adding more material to every prompt.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | microsoft/phi-4 |
| Context window | 16,384 tokens Published checkpoint context; serving engines may impose lower limits Phi-4 official model card |
| Model size | 14B densePhi-4 official model card |
| Input and output | Text input and text outputPhi-4 official model card |
| Weight license | MITPhi-4 official model card |