{
  "id": 14831648,
  "title": "Meta is taking on China’s open AI giants — with a little help from Apple",
  "url": "https://urgent.news/2026/10/16/meta-is-taking-on-chinas-open-ai-giants-with-a-little-help-from-apple",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-16T00:55:10.000Z",
  "source": {
    "name": "Malay Mail",
    "slug": "malay-mail-malaymail",
    "url": "https://www.malaymail.com/news/tech-gadgets/2026/10/16/meta-is-taking-on-chinas-open-ai-giants-with-a-little-help-from-apple/239260"
  },
  "original_language": "en",
  "account": "Meta has entered the competitive arena of open-weight AI models by unveiling Muse Spark Open, a model that the company claims rivals and sometimes surpasses similar products from leading Chinese developers. This new model is the most capable US open-weight model currently available, according to Meta's chief AI officer Alexandr Wang, who shared the news on X.\n\nMuse Spark Open stands out due to its openness, allowing developers to download its parameters or weights and then customize it according to their needs. This contrasts with closed models like Anthropic’s Claude or OpenAI’s ChatGPT, whose functionalities are not customizable by external users. Over the past couple of years, Chinese developers have made significant strides in the open-weight market, with Alibaba’s Qwen, Moonshot AI’s Kimi K3, and Z.ai’s GLM models gaining recognition and competing globally.\n\nMuse Spark Open is designed to perform exceptionally well in coding and agentic work, excelling at long-horizon tasks, complex software engineering, and multi-step reasoning. Meta emphasized that the model includes strong safeguards to ensure safe usage. In addition, the model has been optimized to run on Apple's cutting-edge M5 Ultra chips, marking a unique partnership between Meta and Apple. Meta CEO Mark Zuckerberg tweeted about the seamless performance of Muse Spark Open on Apple's new Mac Studios, crediting Apple's work with CEO John Ternus, who took over as Apple's chief from Tim Cook in September.\n\nWhile Muse Spark Open is a powerful model, it still requires significant computing power. To create a \"supercomputer\" on a single desktop, users need to run the model on two Mac Studios equipped with M5 Ultra chips, according to Meta. This launch follows shortly after Meta's announcement of two new closed models, Wonder and Wonder Full, which the company believes match or even exceed the capabilities of the most advanced versions of ChatGPT and Claude. Meta's pivot to closed models comes as Zuckerberg pushes the company to compete more aggressively with industry leaders OpenAI and Anthropic.\n\nThe AI efforts at Meta initially focused on open models, but recently, Zuckerberg has shifted the company's strategy to compete with OpenAI and Anthropic. The company launched its Spark model in April, a change from its earlier open-source model, Llama, which had not kept pace with competitors in the industry.",
  "summary": "SAN FRANCISCO, Oct 16 — Meta yesterday launched a new open-weight AI model which it claims rivals, and in some cas...",
  "key_points": [
    "Meta launches Muse Spark Open, a top US open-weight AI model.",
    "Model excels in coding and complex tasks, runs on Apple M5 Ultra chips.",
    "Meta shifts strategy to closed models amid competition with OpenAI and Anthropic."
  ],
  "editors_take": "By optimizing its most capable open-weight model for Apple hardware, Meta is positioning itself to challenge Chinese developers and reclaim dominance in the customizable AI market.",
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Financial Times",
        "title": "Meta and Apple team up on AI model tailored for Mac computers",
        "url": "https://urgent.news/2026/10/15/meta-and-apple-team-up-on-ai-model-tailored-for-mac-computers",
        "published": "2026-10-15T20:00:23.000Z"
      }
    ]
  },
  "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."
}