Awesome Smol is a curated list of small, lightweight AI models, tools, and resources for domains like language, audio, vision, and multimodal tasks. These models are designed for edge devices, resource-constrained environments, and rapid prototyping.
Inspiration: awesome-tensorflow-lite.
(Source: Link)
Smol models are AI models optimized for efficiency, offering:
- Lightweight Design: Minimal memory usage.
- Fast Inference: High performance on limited hardware.
- Accessibility: Ideal for mobile, IoT, and edge deployment.
- Versatility: Support across text, audio, vision, and more.
- Highlight community-driven advancements in lightweight AI.
- Provide a central hub for small model resources, tools, and benchmarks.
- Promote faster adoption for real-world applications.
Your contributions are warmly welcome! Submit a pull request (PR) following the contribution guidelines.
- Language Models
- Audio Models
- Vision Models
- Multimodal Models
- Pretrained Models Hub
- Other Insightful Lists
- Tools and Frameworks
- AI Frontier solution at the Edge
- AI News & Announcements
- Resources
| Model | Task | Platform | References |
|---|---|---|---|
| SmolLM | General NLP | Edge, Desktop | Hugging Face |
| Zamba2-7B | Text Understanding | Mobile, Edge | Hugging Face |
| EuroLLM-1.7B | Multilingual NLP | Mobile, IoT | Hugging Face |
| Mistral-Small-Instruct-2409 | Instruction Tasks | Edge, Mobile | Hugging Face |
| Ministral-8B-Instruct-2410 | Instruction Tasks | Edge, Desktop | Hugging Face |
| TinyLlama | Conversational AI | Edge, IoT | Hugging Face |
| Phi-3 | Text Generation | Mobile, Edge | Hugging Face |
| Gemma 2 | Multilingual NLP | Mobile, Desktop | Hugging Face |
| Model | Task | Platform | References |
|---|---|---|---|
| Whisper Small | Speech Recognition | Edge, Desktop | Hugging Face |
| Audio-Mamba (AuM) | Audio Processing | Edge, IoT | GitHub |
| MusicGen | Music Generation | Edge, Desktop | GitHub |
| FastSpeech2 Small | Text-to-Speech | Mobile, Edge | Hugging Face |
| HiFi-GAN Mini | Audio Enhancement | Mobile, IoT | GitHub |
| MatchboxNet Small | Keyword Spotting | Edge, IoT | Hugging Face |
| Model | Task | Platform | References |
|---|---|---|---|
| MobileNet V3 Small | Image Classification | Mobile, Edge | TensorFlow |
| EfficientNet-Lite Small | Image Classification | Mobile, IoT | GitHub |
| YOLOv5 Nano | Object Detection | Edge, IoT | GitHub |
| DeepLab Lite Small | Image Segmentation | Mobile, Edge | GitHub |
| MobileUNet | Image Segmentation | Mobile, IoT | GitHub |
| Vision Transformer Small | Vision Tasks | Mobile, Edge | Hugging Face |
| Model | Task | Platform | References |
|---|---|---|---|
| Mini-DALL-E | Text-to-Image | Mobile, Desktop | GitHub |
| TinyCLIP | Vision-Language | Edge, IoT | Hugging Face |
| Mini-ALIGN | Vision-Language | Mobile, Edge | GitHub |
Pre-trained lightweight models ready for deployment:
- Hugging Face Pretrained Smol Models: Ready-to-deploy smol models with associated datasets, and demo apps (Spaces).
- Model Zoo Models Categories: Open source deep learning code and pretrained models.
- Kaggle Pre-trained Models: Use and download pre-trained models for your machine learning projects.
- Tensorflow Hub: A repository of trained machine learning models.
- Pytorch Hub: Discover and publish models to a pre-trained model repository designed for research exploration.
- edge-ai - @crespum
- awesome-tensorflow-lite - @margaretmz
- Smol Vision - @merveenoyan
- Edge AI Model Zoo - @afondiel
- LiteRT - formerly TensorFlow Lite: Lightweight model deployment for Android.
- CoreML: Apple’s ML framework for iOS.
- ExecuTorch: Pytorch/Meta end-to-end solution for enabling on-device inference.
- ONNX Runtime: Efficient inference engine.
- OpenVINO: Toolkit for optimizing and deploying deep learning models.
- Google AI Edge
- AWS IoT for the Edge
- Azure IoT Edge - Build the intelligent edge
- Qualcomm On-Device AI Solutions
- Meta - Pytorch Edge
- NVIDIA TensorRT
- Edge Impulse
- Edge AI + Vision Alliance
- Edge AI Foundation
[2024/11/26] SmolVLM - small yet mighty Vision Language Model- [2024/09/25] Meta - Llama 3.2: Revolutionizing edge AI and vision with open, customizable models
- [2024/07/16] SmolLM - blazingly fast and remarkably powerful
- [2024/04/23] Introducing Phi-3: Redefining what’s possible with SLMs
- Why Small Language Models (SLMs) Are The Next Big Thing In AI - Forbes (2024/11/25)
- Optimizing Generative AI for Edge Devices
- Deploying ML Models on The Edge - @Microsoft
- Fine-Tuning Small Language Models - Kili
- AI on the edge: latest insights and trends @Qualcomm
- Small is the new big: the rise of small language models - Capgemini
- The 5 leading small language models of 2024: Phi 3, Llama 3, and more - DSDojo
- 7 Steps to Running a Small Language Model on a Local CPU
- SmolLM2 Released: A Series (0.1B, 0.3B, and 1.7B) of Small Language Models for OnDevice Applications
- Machine Learning Systems - Vijay Janapa Reddi / Harvard (online & interactive book)
- Edge-AI Books Collection - @cs-books
- [2024/08/30] Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone - Microsoft
- [2024/06/16] Super Tiny Language Models
- [2024/06/11] Small-E: Small Language Model with Linear Attention for Efficient Speech Synthesis
- [2023/04/28] EDGE IMPULSE: AN MLOPS PLATFORM FOR TINY MACHINE LEARNING