Motif 2 12.7B technical report
Abstract
We introduce Motif-2-12.7B, a new open-weight foundation model that pushes the efficiency frontier of large language models by combining architectural innovation with system-level optimization. Designed for scalable language understanding and robust instruction generalization under constrained compute budgets, Motif-2-12.7B builds upon Motif-2.6B with the integration of Grouped Differential Attention (GDA), which improves representational efficiency by disentangling signal and noise-control attention pathways. The model is pre-trained on 5.5 trillion tokens spanning diverse linguistic, mathematical, scientific, and programming domains using a curriculum-driven data scheduler that gradually changes the data composition ratio. The training system leverages the MuonClip optimizer alongside custom high-performance kernels, including fused PolyNorm activations and the Parallel Muon algorithm, yielding significant throughput and memory efficiency gains in large-scale distributed environments. Post-training employs a three-stage supervised fine-tuning pipeline that successively enhances general instruction adherence, compositional understanding, and linguistic precision. Motif-2-12.7B demonstrates competitive performance across diverse benchmarks, showing that thoughtful architectural scaling and optimized training design can rival the capabilities of much larger models.
Cite
@article{arxiv.2511.07464,
title = {Motif 2 12.7B technical report},
author = {Junghwan Lim and Sungmin Lee and Dongseok Kim and Taehyun Kim and Eunhwan Park and Jeesoo Lee and Jeongdoo Lee and Junhyeok Lee and Wai Ting Cheung and Dahye Choi and Jaeheui Her and Jaeyeon Huh and Hanbin Jung and Changjin Kang and Beomgyu Kim and Minjae Kim and Taewhan Kim and Youngrok Kim and Hyukjin Kweon and Haesol Lee and Kungyu Lee and Dongpin Oh and Yeongjae Park and Bokki Ryu and Dongjoo Weon},
journal= {arXiv preprint arXiv:2511.07464},
year = {2025}
}