ReleaseBytes
gemini Gemini Blog ·

Introducing Gemini 3.8 Flash and 3.8 Flash Cyber AI Models

aigaengineer
feature announcement

Google has introduced Gemini 3.8 Flash, an improved workhorse model for reasoning, coding, and agentic tasks, alongside Gemini 3.8 Flash Cyber, a specialized cybersecurity AI model. Gemini 3.8 Flash offers significant performance gains over its predecessor at the same speed and cost, while 3.8 Flash Cyber delivers frontier-level vulnerability detection and automated patching capabilities. Developers, enterprises, and Google AI subscribers can access Gemini 3.8 Flash, while 3.8 Flash Cyber is available to trusted defenders through the new Fairwind Program. Both models incorporate enhanced safety measures against misuse, with 3.8 Flash Cyber having specific mitigations tailored for cybersecurity experts.

  • →Gemini 3.8 Flash: Enhanced Reasoning and Coding Model
  • →Gemini 3.8 Flash Cyber: Dedicated Cybersecurity AI
  • →Real-world Security Applications and Fairwind Program
  • →Enhanced Safety and Robustness Features
  • →Availability for Developers, Enterprises, and Consumers
Features (2) ›
  • Gemini 3.8 Flash: Enhanced Reasoning and Coding Model

    The Gemini 3.8 Flash model offers significant improvements over 3.7 Flash across software engineering, agentic tasks, and critical multi-step reasoning. It demonstrates strong performance on benchmarks like DeepSWE v1.1 for long-horizon coding and various professional field benchmarks.

  • Gemini 3.8 Flash Cyber: Dedicated Cybersecurity AI

    This variant provides frontier-level performance in autonomous vulnerability discovery and automated patching, outperforming previous models on industry benchmarks like CyberGym and CWE-Bench. It is tailored for trusted defenders and aims to provide a decisive advantage in today’s complex cybersecurity landscape.

Notes (3) ›
  • Real-world Security Applications and Fairwind Program

    Google is already leveraging Gemini 3.8 Flash Cyber internally for securing Chrome and Cloud code, demonstrating its effectiveness in finding and patching vulnerabilities. The model is accessible to trusted government authorities, critical infrastructure operators, and software maintainers through the new Fairwind Program.

  • Enhanced Safety and Robustness Features

    Both Gemini 3.8 models include safeguards against misuse, particularly in CBRN and cyber offense domains, aligning with the Frontier Safety Framework. They also show a significant leap in prompt injection robustness, protecting users from related malicious attacks.

  • Availability for Developers, Enterprises, and Consumers

    Gemini 3.8 Flash is available via the Gemini API, Google AI Studio, Android Studio, and for Google AI Pro/Ultra subscribers across the Gemini app and Google Search. Enterprises can access it in Gemini Enterprise, while cyber professionals can apply for access to 3.8 Flash Cyber through the Fairwind Program.

Read the original announcement →

https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/

Related releases