ZGCM-1: A Fully Open and Extremely Efficient Foundation Model for Math and Agentic Search

ZGCM-1：一个用于数学与智能体搜索的完全开源且极具效率的基础模型

In this work, we present ZGCM-1, a fully open 7B dense foundation model trained from scratch with extreme data, system, and algorithmic efficiency. ZGCM-1 is founded on a core premise: compact models cannot passively memorize the open web, but can overcome parametric capacity limits by coupling deliberate internal thinking with active external tool use. 在这项工作中，我们提出了 ZGCM-1，这是一个完全开源的 7B 稠密基础模型，它通过极高的数据、系统和算法效率从零开始训练。ZGCM-1 基于一个核心前提：紧凑型模型无法被动地记忆整个开放网络，但可以通过将深思熟虑的内部思考与主动的外部工具使用相结合，来克服参数容量的限制。

To support this paradigm across a 256K context, we develop an end-to-end, high-efficiency open training recipe: Architecture & System Co-design: interleaved gated sliding-window and full attention, and a stable FP8 Muon optimizer; Progressive Curriculum & MDP Mid-Training: context scaling across 16K, 64K, and 256K, and the reformulation of interaction traces into Markov Decision Processes. 为了在 256K 上下文中支持这一范式，我们开发了一套端到端、高效率的开源训练方案：架构与系统协同设计（包含交错门控滑动窗口与全注意力机制，以及稳定的 FP8 Muon 优化器）；渐进式课程与 MDP 中期训练（涵盖 16K、64K 和 256K 的上下文扩展，并将交互轨迹重构为马尔可夫决策过程）。

Furthermore, we establish an AI-native R&D workflow where agent swarms autonomously manage cluster operations, data curation, and rapid diagnostic evaluation. Extensive evaluations show that ZGCM-1-7B is competitive across 7B model family on general benchmarks. On several challenging mathematical reasoning and agentic search suites, it remains competitive with frontier models orders of magnitude larger, such as Qwen3-235B-A22B and GLM-5.1. 此外，我们建立了一种 AI 原生的研发工作流，由智能体集群自主管理集群运维、数据整理和快速诊断评估。广泛的评估表明，ZGCM-1-7B 在通用基准测试中与 7B 模型家族相比具有竞争力。在多个极具挑战性的数学推理和智能体搜索套件上，它依然能与规模大出几个数量级的前沿模型（如 Qwen3-235B-A22B 和 GLM-5.1）相媲美。

We also show that our pre-training design offers a ~4.2x efficiency improvement in 16K pre-training time-to-loss. Across the full development lifecycle, we distill eight actionable empirical findings-spanning architectural scaling, SFT quality pruning, long-context generalization, and agentic co-training dynamics. 我们还展示了我们的预训练设计在 16K 预训练的“时间-损失”（time-to-loss）效率上提升了约 4.2 倍。在整个开发生命周期中，我们提炼出了八项可操作的实证发现，涵盖了架构扩展、SFT 质量剪枝、长上下文泛化以及智能体协同训练动态。

To facilitate community research, we open-source model weights from the pre-training, mid-training, and post-training stages, intermediate checkpoints, training code, per-stage data and data recipes, and W&B logs. 为了促进社区研究，我们开源了预训练、中期训练和后训练阶段的模型权重、中间检查点、训练代码、各阶段数据及数据配方，以及 W&B 日志。