{
  "id": 6291563,
  "title": "Latest open artifacts (#24): Motif-3, GLM-5.3, Hy4-preview and open model licenses",
  "url": "https://urgent.news/2026/09/08/latest-open-artifacts-24-motif-3-glm-5-3-hy4-preview-and-open-model",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-08T14:15:25.000Z",
  "source": {
    "name": "Interconnects",
    "slug": "interconnects",
    "url": "https://www.interconnects.ai/p/latest-open-artifacts-24-motif-3"
  },
  "original_language": "en",
  "account": "Motif-3 by Motif-Technologies stands out as one of the hidden gems in the model release scene, showcasing innovation in model training despite limited resources. This new release, building on the success of Motif 2.6B and Motif-2-12.7B, exhibits impressive scores given its size. Meanwhile, RedNote/Xiaohongshu, the Chinese Instagram, demonstrates growing seriousness in model training, having released models as early as 2025 and winning the IMO 2026 with a perfect score using an internal harness. Qwen is teasing the next version of its models, Qwen3.8-Flash-Next, featuring an architecture with 125B-A6B incorporating 51B n-gram embeddings, employing GDN and Qwen Sparse Attention. This preview hints at a growing popularity of similar architectures, with the ecosystem expected to develop integrations in the future. GLM-5.3-Flash by zai-org is the latest addition to the open model licenses, having transitioned from MIT (GLM-5.2 and earlier) to a custom license. The new license requires inference and fine-tuning providers to pass a security review by Z.AI if their aggregate revenue, including affiliates, exceeds 10 billion US dollars within a 12-month period. The ambiguity surrounding the definition of \"affiliates\" adds uncertainty and potential barriers to adoption, highlighting the challenges of creating such licenses in a landscape with numerous open and closed alternatives.",
  "summary": "The open model ecosystem continues to expand in its breadth",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}