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

arXiv:2609.13356v1 Announce Type: new Abstract: 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. 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. 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. 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. To facilitate community research, we open-s

ZGCM-1's efficiency and competitiveness are notable, but implications for human flourishing are unclear.

The ZGCM-1 model demonstrates significant efficiency and capability advancements for math reasoning and agentic search, which can enhance research and automation productivity. However, its agentic nature and autonomous management raise risks of misuse or unintended autonomous behavior. The open-source approach promotes transparency but could facilitate misuse by malicious actors.

This is a technical efficiency and openness advance, not a claim about societal harm or benefit beyond the usual dual-use dynamics of capable open models. The open-sourcing and efficiency gains are mildly positive for access and reproducibility, tempered by reduced human oversight in autonomous training workflows and unverified preprint status.

The model demonstrates significant efficiency improvements and competitive performance in mathematical reasoning and agentic search, making advanced AI capabilities more accessible. The open-sourcing of model weights and training details fosters collaborative research and innovation. This accelerates AI development, which has both potential benefits and risks for society.

SOURCE     arXiv cs.AI (tier 1)
   ↓
DOCUMENT   60f6cbca-23d6-4a15-85fd-e59df2835c7b
           https://arxiv.org/abs/2609.13356
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  4 runs, methodology 1.0.1
   ↓
SCORE      -14  (Mixed or uncertain)
   ↓
CONFIDENCE 69%