{
  "id": 5114812,
  "title": "Introducing Gemini 3.8 Flash and 3.8 Flash Cyber",
  "url": "https://urgent.news/2026/09/02/introducing-gemini-3-8-flash-and-3-8-flash-cyber",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-02T15:00:00.000Z",
  "source": {
    "name": "Google Blog",
    "slug": "google-blog",
    "url": "https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/"
  },
  "original_language": "en",
  "account": "Google has unveiled Gemini 3.8 Flash and 3.8 Flash Cyber, the latest iterations of their best reasoning and coding models, boasting speed and low cost similar to their 3.7 counterparts. These two variants, though designed for distinct deployment environments, share the same foundational intelligence and are bolstered by sophisticated agentic loops that refine underlying models through recursive evaluation.\n\nA key driver behind the enhancements in Gemini 3.8 Flash is rigorous training in the cybersecurity domain. In particular, 3.8 Flash excels in solving intricate software engineering problems autonomously, outperforming most larger frontier models on DeepSWE v1.1, all while costing less. The model's dependability in critical enterprise autonomy across specialized fields is also noteworthy. In quantitative and professional domains that demand advanced analysis and reporting, 3.8 Flash outshines 3.7 Flash and other frontier models in benchmark tests such as Vals Finance Agent V2 and Harvey's Legal Agent Benchmark.\n\nNotably, 3.8 Flash demonstrates a heightened level of diligence on complex tasks, executing additional reasoning steps and invoking tools iteratively. This extra effort may result in more token usage, particularly at higher effort levels. However, developers can choose lower effort levels for cost efficiency or continue using Gemini 3.7 Flash for efficiency-focused workloads. Gemini 3.8 Flash Cyber, available to a select group of trusted defenders through the Fairwind Program, offers a significant edge in the rapidly evolving cybersecurity landscape, delivering Flash speed and cost-effective performance in autonomous vulnerability discovery.\n\nOn the CyberGym benchmark, Gemini 3.8 Flash Cyber exhibits frontier-level performance in autonomously discovering vulnerabilities, surpassing both 3.5 Flash Cyber and larger frontier models. The model's success rate in an extensive internal benchmark that tests its ability to identify vulnerabilities across 20 programming languages exceeds 70%. Gemini 3.8 Flash Cyber focuses on empowering defenders with expert capabilities, prioritizing vulnerability fixing over offensive techniques like exploitation. This model is already being utilized by Google to secure code, featuring safeguards against misuse in CBRN and cyber offense domains while enabling beneficial use cases, as per the Frontier Safety Framework.",
  "summary": "Gemini 3.8 Flash and 3.8 Flash Cyber deliver next-generation intelligence for agentic workflows and cybersecurity.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 2,
    "also_reported_by": [
      {
        "outlet": "Google DeepMind",
        "title": "Introducing Gemini 3.8 Flash and 3.8 Flash Cyber",
        "url": "https://urgent.news/2026/09/02/introducing-gemini-3-8-flash-and-3-8-flash-cyber-5122354",
        "published": "2026-09-02T16:18:31.000Z"
      }
    ]
  },
  "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."
}