{
  "id": 17074645,
  "title": "Mistral says its new model, Large 4, achieves ‘frontier’ performance among open models for agentic coding and cyber defense",
  "url": "https://urgent.news/2026/10/29/mistral-says-its-new-model-large-4-achieves-frontier-performance",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-29T14:00:00.000Z",
  "source": {
    "name": "Fortune",
    "slug": "fortune",
    "url": "https://fortune.com/2026/10/29/mistral-releases-new-model-large-4-and-claims-top-open-model-for-agentic-coding-and-cyber-defense/"
  },
  "original_language": "en",
  "account": "French AI startup Mistral has unveiled its latest model, Mistral Large 4, asserting it has edged past Chinese competitors to the forefront of open-weight AI technology. Unlike most leading AI models, open-weight models are freely accessible for public use, modification, and deployment. Mistral claims its new 1 trillion-parameter model, available on the open-source platform Hugging Face, is the most proficient open model developed in the West. The startup asserts that Large 4 narrowly surpasses all other open models on Artificial Analysis’ Coding Agent Index, an independent benchmark evaluating AI agents' proficiency in executing complete software engineering tasks. Mistral's model ties with China's GLM-5.3 on Terminal-Bench 4.0, a coding agent test, following reinforcement learning training since its preview release on October 6. The company's performance on Terminal Bench 4.0 improved from 28% accuracy to 41% after additional reinforcement learning training. Despite OpenAI and Anthropic's proprietary models outperforming the benchmark on overall capabilities, such as Anthropic's Mythos 5.1 scoring 60.9%, Large 4 proves to be more efficient, consuming 47% fewer tokens per task compared to GLM-5.3. Mistral often refers to the model as \"Le Chonk,\" a playful nod to the company's cat mascot. Mistral VP of science Pierre Stock attributes the gains to three weeks of reinforcement learning, highlighting the release as a \"pivotal moment in the global AI race\" and \"a turning point for Europe,\" enabling the continent to develop an ecosystem around AI foundation models. Stock emphasized that Large 4 excels in cyber defense, arguing that closed models may deny or limit access based on perceived offensive cyber tasks. Mistral's model is now considered the default choice for businesses, particularly in regulated sectors, who are concerned about potential access cuts to closed models.",
  "summary": "The French startup says its new trillion-parameter model is the best open model built in the West, and is pitching it to businesses worried about losing access to closed AI.",
  "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."
}