Every Activation Boosted: Scaling General Reasoner to 1 Trillion Open Language Foundation
Abstract
We introduce Ling 2.0, a series reasoning-oriented language foundation built upon the principle that every activation boosts reasoning capability. Designed to scale from tens of billions to one trillion parameters under a unified Mixture-of-Experts (MoE) paradigm, Ling 2.0 emphasizes high sparsity, cross-scale consistency, and efficiency guided by empirical scaling laws. The series includes three non-thinking (instruct) models - Ling-mini-2.0, Ling-flash-2.0, and Ling-1T - ranging from 16B to 1T total parameters and achieving up to 7-fold active-compute efficiency compared with dense counterparts. Ling 2.0 integrates coordinated innovations across model architecture, pre-training, post-training, and infrastructure: a high-sparsity MoE with MTP for efficient reasoning, reasoning-oriented data and mid-training CoT activation, reinforcement-based fine-tuning (DFT, Evo-CoT), and full-scale FP8 training with fine-grained heterogeneous pipelines. At the trillion scale, Ling-1T establishes a new Pareto frontier of reasoning accuracy versus computational efficiency, demonstrating that sparse activation, when properly aligned with reasoning objectives, enables scalable and efficient intelligence. Collectively, Ling 2.0 provides a coherent, open, and efficient foundation for advancing future reasoning and thinking models, including the Ring series built upon the same base.
Keywords
Cite
@article{arxiv.2510.22115,
title = {Every Activation Boosted: Scaling General Reasoner to 1 Trillion Open Language Foundation},
author = {Ling Team and Ang Li and Ben Liu and Binbin Hu and Bing Li and Bingwei Zeng and Borui Ye and Caizhi Tang and Changxin Tian and Chao Huang and Chao Zhang and Chen Qian and Chenchen Ju and Chenchen Li and Chengfu Tang and Chilin Fu and Chunshao Ren and Chunwei Wu and Cong Zhang and Cunyin Peng and Dafeng Xu and Daixin Wang and Dalong Zhang and Dingnan Jin and Dingyuan Zhu and Dongke Hu and Fangzheng Zhao and Feifan Wu and Feng Zhu and Gangshan Wang and Haitao Zhang and Hailin Zhao and Hanxiao Zhang and Hanzi Wang and Hao Qian and Haoyi Yu and Heng Zhang and Hongliang Zhang and Hongzhi Luan and Huirong Dong and Huizhong Li and Jia Li and Jia Liu and Jialong Zhu and Jian Sha and Jianping Wei and Jiaolong Yang and Jieyue Ma and Jiewei Wu and Jinjing Huang and Jingyun Tian and Jingyuan Zhang and Jinquan Sun and Juanhui Tu and Jun Liu and Jun Xu and Jun Zhou and Junjie Ou and Junpeng Fang and Kaihong Zhang and Kaiqin Hu and Ke Shi and Kun Tang and Kunlong Chen and Lanyin Mei and Lei Liang and Lei Xu and Libo Zhang and Lin Ju and Lin Yuan and Ling Zhong and Lintao Ma and Lu Liu and Lu Yu and Lun Cai and Meiqi Zhu and Mengying Li and Min Chen and Minghao Xue and Minghong Cai and Mingming Yin and Peijie Jiang and Peilong Zhao and Pingping Liu and Qian Zhao and Qing Cui and Qingxiang Huang and Qingyuan Yang and Quankun Yu and Shaowei Wei and Shijie Lian and Shoujian Zheng and Shun Song and Shungen Zhang and Shuo Zhang and Siyuan Li and Song Liu and Ting Guo and Tong Zhao and Wanli Gu and Weichang Wu and Weiguang Han and Wenjing Fang and Wubin Wang and Xiang Shu and Xiao Shi and Xiaoshun Lan and Xiaolu Zhang and Xiaqing Sun and Xin Zhao and Xingyu Lu and Xiong Xu and Xudong Wang and Xudong Wang and Xuemin Yang and Yajie Yang and Yang Xiang and Yanzhe Li and Yi Zhang and Yilong Wang and Yingxue Li and Yongzhen Guo and Yuzhuo Fu and Yuanyuan Wang and Yue Yang and Yue Yu and Yufeng Deng and Yun Zhang and Yunfei Yu and Yuqi Zhang and Yuxiao He and Zengke Gui and Zhaoxin Huan and Zhaoyang Wang and Zhibo Zhu and Zhihao Wang and Zhiqiang Zhang and Zhoufei Wang and Zihang Zeng and Ziqi Liu and Zitao Xuan and Zuoli Tang},
journal= {arXiv preprint arXiv:2510.22115},
year = {2025}
}
Comments
Ling 2.0 Technical Report