Agents-A1：35B MoE 智能体模型通过扩展 horizon 达到万亿参数级性能
研究人员提出 Agents-A1，一个 35B 参数的 Mixture-of-Experts 智能体模型，通过扩展智能体 horizon（长轨迹与异构能力两个视角）达到万亿参数模型性能。团队构建了长 horizon 知识-行动基础设施，生成平均 45K token 的智能体轨迹，并采用三阶段训练：全领域监督微调、领域级教师模型训练、多教师领域路由在线蒸馏（含显著词汇对齐）。对比万亿参数模型 Kimi-K2.6 和 DeepSeek-V4-pro，Agents-A1 在 SEAL-0（56.4）、IFBench（80.6）、HiPhO（46.4）、FrontierScience-Olympiad（79.0）和 MolBench-Bind（56.8）上领先，并在 SciCode（44.3）、HLE（47.6）和 BrowseComp（75.5）上保持强竞争力。
用35B模型追平1T参数模型，这条“扩展智能体视野”的路比无脑堆参数务实得多，做Agent和长程推理的团队必须认真读。
Authors:Lei Bai, Zongsheng Cao, Yang Chen, Zhiyao Cui, Shangheng Du, Yue Fan, Shiyang Feng, Zijie Guo, Haonan He, Liang He, Xiaohan He, Shuyue Hu, Yusong Hu, Songtao Huang, Yichen Jiang, Hao Li, Xin Li, Dahua Lin, Weihao Lin, Fenghua Ling, Dongrui Liu, Zhuo Liu, Wenjie Lou, Runmin Ma, Chunjiang Mu, Haoyang Peng, Tianshuo Peng, Jinxin Shi, Luohe Shi, Boyuan Sun, Zelin Tan, Shengji Tang, Yan Teng, Qianyi Wang, Xiaosong Wang, Yiming Wu, Yi Xie, Xiangchao Yan, Jingqi Ye, Peng Ye, Fangchen Yu, Jiakang Yuan, Bihao Zhan, Bo Zhang, Chen Zhang, Shufei Zhang, Shuaiyu Zhang, Wenlong Zhang, Yiqun Zhang, Junpeng Zhao, Zhijie Zhong, Bowen Zhou, Yuhao Zhou
Abstract:We introduce Agents-A1, a 35B Mixture-of-Experts Agentic Model that reaches trillion-parameter-level performance by scaling the agent horizon. We investigate agent-horizon scaling from two perspectives: scaling long-horizon trajectories and scaling heterogeneous agent abilities. To support this goal, we build a long-horizon knowledge-action infrastructure that connects external knowledge, actions, observations, and verifier outcomes, producing agentic trajectories with an average length of 45K tokens. Based on this, we train Agents-A1 with a three-stage recipe. First, we perform full-domain supervised fine-tuning to align the base model with broad agentic behaviors. Second, we train domain-level teacher models to capture specialized expertise in each domain. Third, we propose a multi-teacher domain-routed on-policy distillation with salient vocabulary alignment to improve knowledge transfer efficiency across different domains, unifying six heterogeneous domains into one deployable student model. Agents-A1 achieves strong and broad performance for long-horizon agent benchmarks. Compared with 1T-parameter model such as Kimi-K2.6 and DeepSeek-V4-pro, Agents-A1 achieves leading results on SEAL-0 (56.4), IFBench (80.6), HiPhO (46.4), FrontierScience-Olympiad (79.0), and MolBench-Bind (56.8), and remains highly competitive on SciCode (44.3), HLE (47.6) and BrowseComp (75.5). We hope this work provides the community with a practical path for scaling the horizon using a 35B agent that can reach or match the performance of 1T models on long-horizon tasks.
| Comments: | The model checkpoints and evaluation codebase are available at this https URL and this https URL |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2606.30616 [cs.CL] |
| (or arXiv:2606.30616v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2606.30616 arXiv-issued DOI via DataCite |
Submission history
From: Bo Zhang [view email]
[v1]
Mon, 29 Jun 2026 17:50:54 UTC (12,414 KB)
[v2]
Mon, 13 Jul 2026 13:01:42 UTC (12,414 KB)
来源：HuggingFace Daily Papers（社区热门论文） · arxiv.org