Show HN: Sweep, Open-weights 1.5B model for next-edit autocomplete (huggingface.co)

A new AI tool called Sweep has been introduced, featuring a 1.5 billion parameter model designed for next-edit autocomplete in coding. Quantized in the Q8_0 GGUF format, Sweep effectively predicts your next code modification before you even make it, achieving remarkable performance on next-edit benchmarks—surpassing larger models by over four times. This efficiency enables it to run locally on personal laptops in under 500 milliseconds, thanks to the use of speculative decoding. The significance of Sweep lies in its ability to enhance coding productivity, making it an appealing development for the AI and machine learning community. With a context length of 8,192 tokens and built on the Qwen2.5-Coder as its base model, Sweep utilizes a specialized prompt format that incorporates file context, recent changes, and the current state of coding projects to anticipate user edits accurately. Those interested can easily deploy the model using straightforward installation commands, fostering greater accessibility in AI-driven coding tools.