Sweep Next-Edit 1.5B (GGUF)

Sweep Next-Edit 1.5B (GGUF)

Predicts your next code edit from the open file, recent diffs and current state, running locally as a 1.5B Q8_0 GGUF model. For developers who want edit autocomplete on their own laptop.

Made by sweepai on Hugging Face. These notes are Mindket's, written from its model page and model card; the model's own files and license are on Hugging Face.

About

What it is: Sweep Next-Edit 1.5B is a small model for next-edit autocomplete, stored as a Q8_0 GGUF file. Instead of finishing the word you are typing, it guesses which edit you will make next. The card says it runs on a laptop in under 500 ms with speculative decoding, and beats models more than four times larger on next-edit benchmarks.

Who it suits: It suits developers who want edit suggestions that run on their own machine, and people building editor tools around a local model.

Industry | Who uses it | How they use it | Example
Software development | Individual developers | Run next-edit suggestions locally while coding | Seeing the likely follow-up change after you rename a variable
Developer tooling | Editor plugin authors | Build a local suggestion feature around the GGUF model | Wiring the model into an editor extension
AI research | Code model researchers | Study a small model tuned for edit prediction | Testing it on your own recorded edits

You can skip it if you want a general chat model, because this one only predicts edits.

Where to use it:

• A Python script that uses llama-cpp-python, as the card's usage shows
• Your laptop, with Windows, macOS or Linux and Python
• Editors, through tools you build yourself, since a thread linked from the card discusses VSCode, Neovim and Emacs

Use case | Who | What they get back
Next-edit autocomplete | Developers | A predicted next change to your code
Local suggestions | Privacy-minded teams | Predictions made on your own machine
Editor integration | Tool builders | A small model to embed in an editor workflow

How to set it up:

• Python and the uv installer
• The huggingface_hub package
• The model file and the run_model.py script from the repository
• llama-cpp-python, to run the GGUF file

1. Install the Hugging Face hub package.
2. Download the repository files, which hold the model file and run_model.py.
3. Install llama-cpp-python and huggingface_hub with uv.
4. Run the script.

pip install -U huggingface_hub
hf download sweepai/sweep-next-edit-1.5B

uv pip install llama-cpp-python huggingface_hub
python run_model.py

How to use it: You run the script from the card with Python. It builds the model's prompt from your file, your recent changes and the current state, then asks the model for the next edit.

Tool | What it does
run_model.py | The card's full example of building the prompt and running the model
llama-cpp-python | Loads and runs the GGUF model
huggingface_hub | Fetches the model files

Example: You have a file open and have just changed a line. The script packs that file, your recent diffs and the current state into the model's prompt format. The model returns the edit it expects you to make next.

For newcomers and experts: Next-edit autocomplete suggests the whole change you are likely to make, not just the next word. To start, download the files, install the packages, and run run_model.py.

If you know the field, this is a 1.5B model on a Qwen2.5-Coder base, quantized to Q8_0 and served through llama.cpp bindings. Its context length is 8192 tokens. It expects one fixed prompt layout, so you must follow the example script closely.

Your file and recent diffs are collected
They are packed into the model's prompt format
The model runs locally through llama-cpp-python
It returns the predicted next edit

Good to know:

• Format: GGUF with Q8_0 quantization
• Context length: 8192 tokens
• Base model: Qwen2.5-Coder
• Speed and benchmark claims come from the authors, and the card shows no test numbers
• License: Apache 2.0
• Cost to run: none named, it runs on your own laptop
• Data it reaches: the code files and diffs you feed it, and the files you download from Hugging Face
• Not for: free-form chat, since it expects a fixed prompt layout

On Hugging Face, 13 November 2026

• Parameters: 1.4B
• Weights: about 723 MB at 4-bit (estimate)
• Formats: GGUF
• Downloads in the last 30 days: 1,142
• Likes: 341
• License: apache-2.0
• Quantized versions: 2
• Fine-tunes: 1
• Tags: gguf, code, autocomplete, next-edit

Links

• Model: https://huggingface.co/sweepai/sweep-next-edit-1.5B
• Files: https://huggingface.co/sweepai/sweep-next-edit-1.5B/tree/main

Get the model

Download its files with Hugging Face's command line:

pip install -U huggingface_hub
hf download sweepai/sweep-next-edit-1.5B

Then follow the usage in its model card. Nothing in this kit runs by itself.