Hugging Face has introduced SmolTools, a set of applications built on the recently launched SmolLM2 model, a compact 1.7-billion parameter language model. SmolTools includes specialized tools for summarization, rewriting, and task automation, bringing efficient AI functionality to a broader range of users.
SmolTools suite includes several applications designed to streamline common tasks:
- SmolSummarizer: Enables quick summarization for texts up to 20 pages, retaining key points and supporting follow-up questions for deeper understanding.
- SmolRewriter: Refines initial drafts to sound professional and approachable while preserving original intent, ideal for email and messaging needs.
- SmolAgent: Acts as a tool-integrated AI agent capable of executing tasks like random number generation or time checks. Its extensible tool system also allows users to add new capabilities as needed.
To install SmolTools, users can follow these setup steps:
1. Clone the repository:
git clone https://github.com/huggingface/smollm.git
cd smollm/smol_tools
2. Install dependencies:
uv venv --python 3.11
source .venv/bin/activate
uv pip install -r requirements.txt
These tools are powered by SmolLM2’s variants, including lighter models (360M and 135M), optimized for devices with limited resources. This development brings AI-powered functions to a wider range of platforms, with implications for small businesses, developers, and edge devices.
Drasko Draskovic noted the potential impact:
For small businesses, individual developers, and even edge devices like smartphones, this is game-changing. Imagine running sophisticated summarization or rewriting tasks directly on-device, empowering users everywhere with AI that’s accessible, efficient, and practical.
By pushing forward with innovations like SmolTools, Hugging Face is not just developing technology. They are helping democratize AI. They are proving that efficiency and accessibility are as important as power, opening doors to a future where AI is integrated into everyday workflows, making an impact on all levels of business and society.
SmolLM2’s on-device performance is enhanced with support for tool calling and structured outputs, features critical for building advanced workflows and agentic AI applications. Gaurav Dhiman raised the importance of these functions:
Without that, it is practically not possible to build useful AI apps other than general chatting summarization apps. For building something serious like Agentic workflows, both tool calling and structured outputs are crucial capabilities.
Andrés Marafioti, a machine learning researcher at Hugging Face, confirmed SmolTools support for these features, referencing a repository example that includes an agent for function calling and structured outputs.
SmolTools offers accessible, practical tools that simplify text processing tasks on-device, with potential applications across various fields.
