{
  "id": 3702881,
  "title": "Zhipu AI shares jump as viral Ox Alpha model revealed as GLM-5.3-Flash on Chinese chips",
  "url": "https://urgent.news/2026/08/27/zhipu-ai-shares-jump-as-viral-ox-alpha-model-revealed-as-glm-5-3-3702881",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-27T08:30:12.000Z",
  "source": {
    "name": "SCMP Business",
    "slug": "scmp-business",
    "url": "https://www.scmp.com/tech/big-tech/article/3365433/zhipu-ai-shares-jump-viral-ox-alpha-model-revealed-glm-53-flash-chinese-chips"
  },
  "original_language": "en",
  "account": "China's Zhipu AI introduced its newest open-weight model, GLM-5.3-Flash, formerly known as Ox Alpha, stating that it ran entirely on a cluster of 100,000 domestically produced chips during a high-profile stealth trial. Following a week of heightened activity on AI model marketplace OpenRouter and agent platform OpenCode, the model processed 62 trillion tokens before its official launch on Wednesday, according to Zhipu. On OpenRouter, the system processed over 11 trillion tokens in the first three days, making it the platform's biggest launch to date. Zhipu's shares increased by more than 12 percent to HK$1,160 in Hong Kong on Thursday. This deployment signifies a significant test of China's capacity to manage large-scale global inference workloads on locally manufactured hardware, as Beijing aims to lessen its dependence on advanced processors from US leader Nvidia due to stringent export restrictions. Upon its preview, Ox Alpha quickly became the top-ranked model on global usage charts, processing 10.3 trillion tokens, or nearly 31 percent of the platform's total weekly volume, based on OpenRouter data on Thursday. To tackle the limited memory capacity and bandwidth of individual Chinese chips compared to leading Nvidia graphics processing units, Zhipu, which operates internationally under the Z.ai brand, developed a specialized inference engine that divided processing stages into separately managed computing pools, tripling overall serving performance from its initial baseline and achieving hardware efficiency and per-token costs comparable to mainstream Nvidia accelerators. While Zhipu did not disclose the specific chip suppliers for this cluster, it has previously worked with prominent domestic semiconductor firms, including Huawei Technologies, Cambricon Technologies, and Moore Threads. Cambricon announced that it had achieved \"Day 0\" compatibility to serve GLM-5.3-Flash. GLM-5.3-Flash, with 320 billion total parameters, activated only 18 billion per request to minimize computing overhead, and was the first model in the GLM-5 series to process visual information along with text. Benchmarking firm Artificial Analysis assigned the model a score of 57 on its Intelligence Index, positioning it 10th globally and third among open-weight models, trailing Moonshot AI's Kimi K3 and Alibaba Group Holding's Qwen3.8 2.4T A95B. Zhipu is promoting competitive pricing to attract international developers, offering GLM-5.3-Flash at one-tenth the price of standard GLM-5.3 – dropping to one-twentieth under a limited promotion – claiming the new model costs approximately one-fortieth as much as Anthropic's Opus 4.8 at equivalent intelligence levels. Despite high traffic during the free trial, early developer feedback was mixed, with users praising the model's capacity to debug complex code (solving 28 percent of 175 LiveCodeBench problems) but also reporting occasional hallucinations, dropped tasks, and sluggish generation. Artificial Analysis noted that GLM-5.3-Flash's output speed trailed the industry average. Zhipu has released the model weights globally and incorporated GLM-5.3-Flash across its application programming interface, ZCode platform, and GLM Coding Plan. The launch aligns with heightened competition in China's open-source landscape. On Wednesday, Alibaba unveiled Qwen3.8-Flash-Next, a multimodal preview of Qwen4 that reportedly activated 6 billion of its 125 billion parameters to similarly reduce inference costs. Alibaba owns the South China Morning Post.",
  "summary": "China’s Zhipu AI has launched its latest open-weight model, GLM-5.3-Flash – previously code-named Ox Alpha – saying that the system ran entirely on a cluster of 100,000 domestically produced chips during a high-profile stealth trial. The announcement followed a week of heavy traffic on artificial intelligence model marketplace OpenRouter and agent platform OpenCode, where the model processed 62…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 6,
    "also_reported_by": [
      {
        "outlet": "Techmeme",
        "title": "Alibaba releases Qwen3.8-Flash, an open-weight, 125B-parameter model built on its next-gen Qwen 4 architecture, saying it rivals Opus 4.6 and V4-Flash (Luz Ding/Bloomberg)",
        "url": "https://urgent.news/2026/08/26/alibaba-releases-qwen3-8-flash-an-open-weight-125b-parameter-model",
        "published": "2026-08-26T12:51:34.000Z"
      },
      {
        "outlet": "Techmeme",
        "title": "Z.ai releases GLM-5.3-Flash, the first natively multimodal GLM-5 series model, with 320B parameters, saying it outperforms GLM-5.2 at \"one-tenth the price\" (Z.ai)",
        "url": "https://urgent.news/2026/08/26/z-ai-releases-glm-5-3-flash-the-first-natively-multimodal-glm-5",
        "published": "2026-08-26T14:26:48.000Z"
      },
      {
        "outlet": "Economic Times Tech",
        "title": "Alibaba's Qwen launches Qwen3.8-Flash AI model with lower training costs",
        "url": "https://urgent.news/2026/08/26/alibabas-qwen-launches-qwen3-8-flash-ai-model-with-lower-training",
        "published": "2026-08-26T15:02:52.000Z"
      },
      {
        "outlet": "SiliconANGLE",
        "title": "Z.ai open-sources ‘Ox Alpha’ model as GLM-5.3-Flash",
        "url": "https://urgent.news/2026/08/27/z-ai-open-sources-ox-alpha-model-as-glm-5-3-flash",
        "published": "2026-08-27T00:26:24.000Z"
      },
      {
        "outlet": "SCMP Tech",
        "title": "Zhipu AI shares jump as viral Ox Alpha model revealed as GLM-5.3-Flash on Chinese chips",
        "url": "https://urgent.news/2026/08/27/zhipu-ai-shares-jump-as-viral-ox-alpha-model-revealed-as-glm-5-3",
        "published": "2026-08-27T08:30:12.000Z"
      }
    ]
  },
  "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."
}