Overview
Xenova/sweep-next-edit-1.5B is a compact yet powerful 1.5 billion parameter model built on the Qwen2.5-Coder architecture. Developed by Xenova, its core function is next-edit autocomplete, predicting the subsequent code modification a user intends to make. This model is engineered for local execution, achieving sub-500ms response times through speculative decoding, making it highly efficient for real-time coding assistance.
Key Capabilities
- Next-Edit Autocomplete: Specializes in predicting the next logical code edit within a development environment.
- High Performance: Outperforms models significantly larger in size on specific next-edit benchmarks.
- Local Execution: Designed to run efficiently on local machines, ensuring fast response times.
- Code Context Understanding: Utilizes a sophisticated prompt format to incorporate context files, recent diffs, and current file states for accurate predictions.
Model Details
- Parameters: 1.5 billion
- Base Model: Qwen2.5-Coder
- Context Length: 8192 tokens (though the model card states 131072, the README specifies 8192).
Good For
- Developers seeking highly efficient and accurate code autocomplete features.
- Integrating into IDEs or code editors for real-time, local code suggestions.
- Enhancing developer workflow by predicting and suggesting code changes before they are manually typed.