Small audio language models achieve strong on-device speech understanding
Samsone: A Family of Open Small Audio Language Models for On-Device Inference
Artificial Intelligence
Summary
Big audio language models are powerful but usually need large computers to work, raising privacy and speed issues. The authors created Samsone, a set of smaller audio language models that run directly on devices like phones. Samsone models can handle audio tasks quickly without needing internet connections. Despite being smaller, these models perform well compared to much bigger ones. The authors share their models, code, and an app so others can use and build on their work.
What this means in practice
- •For mobile app developers: Integrate Samsone models to run speech understanding directly on phones without cloud processing, reducing latency and improving user privacy.
- •For wearable device engineers: Embed compact Samsone models in wearables to provide real-time audio language processing with minimal hardware demands.
Authors
Piotr Masztalski, Michał K. Grzeszczyk, Olaf Sikorski
Abstract
The success of Large Audio Language Models has driven the development of massive multimodal networks exceeding billions of parameters. However, the demand for privacy-preserving, low-latency processing has shifted focus toward Small Audio Language Models (SALMs) capable of on-device execution. In this paper, we introduce Samsone, a family of SALMs designed for edge computing. Our core model, Samsone-134M, establishes a new state-of-the-art for its size class across multiple benchmarks. We further explore the scaling laws of SALMs by introducing Samsone-99M and Samsone-356M. Despite their compact footprint, the Samsone family delivers performance competitive with models orders of magnitude larger. To foster open research and reproducibility, we train Samsone on publicly available data. We release the training code, model weights, mobile-optimized checkpoints and provide an open-source Android application to demonstrate real-time on-device inference of Samsone.