{
  "id": 12424385,
  "title": "Mistral Large 4",
  "url": "https://urgent.news/2026/10/06/mistral-large-4-12424385",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-06T13:15:49.000Z",
  "source": {
    "name": "Hacker News",
    "slug": "hacker-news",
    "url": "https://mistral.ai/news/mistral-large-4//"
  },
  "original_language": "en",
  "account": "Mistral Large 4, unofficially referred to as ML4, has been officially launched as a public preview. This 1 trillion-parameter, natively multimodal model boasts 49 billion active parameters, marking it as Mistral's largest and most capable model to date. ML4 demonstrates exceptional performance across various domains, including coding, agentic workflows, and multimodal understanding.\n\nThe model's performance surpasses that of other open-weight models developed globally, with some capabilities even surpassing closed models. It has been trained on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's European datacenters. The public preview is accessible on Mistral Studio using the same infrastructure.\n\nML4 was trained from scratch, using a diverse multilingual dataset spanning over 160 languages, including every official EU language. The model's training involved leading enterprises across various industries, including finance, engineering, manufacturing, logistics, pharmaceuticals, public sector, and more.\n\nML4's strengths lie in its ability to handle complex software engineering tasks, repository understanding, and complex terminal workflows, placing it ahead of several other models. It also exhibits strong capabilities in image understanding and reasoning across complex documents, charts, and professional deliverables like spreadsheets, slides, and PDFs.\n\nThe model's development was driven by Mistral's long-term investment in infrastructure, research, and product development. It is designed to provide customers with control over their AI, particularly in critical sectors like cybersecurity, where provider-level refusals can hinder legitimate vulnerability research and incident response.",
  "summary": null,
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 3,
    "also_reported_by": [
      {
        "outlet": "Hacker News",
        "title": "Mistral Large 4",
        "url": "https://urgent.news/2026/10/06/mistral-large-4",
        "published": "2026-10-06T13:15:49.000Z"
      },
      {
        "outlet": "Hacker News Best",
        "title": "Mistral Large 4",
        "url": "https://urgent.news/2026/10/06/mistral-large-4-12432452",
        "published": "2026-10-06T13:15:49.000Z"
      }
    ]
  },
  "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."
}