We present Fox-1, a series of small language models (SLMs) consisting of Fox-1-1.6B and Fox-1-1.6B-Instruct-v0.1. These models are pre-trained on 3 trillion tokens of web-scraped document data and fine-tuned with 5 billion tokens of instruction-following and multi-turn conversation data. Aiming to improve the pre-training efficiency, Fox-1-1.6B model introduces a novel 3-stage data curriculum across all the training data with 2K-8K sequence length. In architecture design, Fox-1 features a deeper layer structure, an expanded vocabulary, and utilizes Grouped Query Attention (GQA), offering a performant and efficient architecture compared to other SLMs. Fox-1 achieves better or on-par performance in various benchmarks compared to StableLM-2-1.6B, Gemma-2B, Qwen1.5-1.8B, and OpenELM1.1B, with competitive inference speed and throughput. The model weights have been released under the Apache 2.0 license, where we aim to promote the democratization of LLMs and make them fully accessible to the whole open-source community.
@article{arxiv.2411.05281,
title = {Fox-1: Open Small Language Model for Cloud and Edge},
author = {Zijian Hu and Jipeng Zhang and Rui Pan and Zhaozhuo Xu and Shanshan Han and Han Jin and Alay Dilipbhai Shah and Dimitris Stripelis and Yuhang Yao and Salman Avestimehr and Tong Zhang and Chaoyang He},
journal= {arXiv preprint arXiv:2411.05281},
year = {2025}
}
Comments
Base model is available at https://huggingface.co/tensoropera/Fox-1-1.6B and the instruction-tuned version is available at https://huggingface.co/tensoropera/Fox-1-1.6B-Instruct-v0.1