Google announces Gemini 3.8 Flash and 3.8 Flash Cyber AI models

Google has unveiled Gemini 3.8, a cutting-edge AI model featuring two versions: Gemini 3.8 Flash and Gemini 3.8 Flash Cyber. The company describes Gemini 3.8 Flash as their "most intelligent workhorse model" that excels in software engineering, agentic tasks, and specialized domain reasoning. This model will be accessible at the same introductory price as its predecessor, 3.7 Flash, costing $0.75 per million input tokens and $3.75 per million output tokens.

In contrast, Gemini 3.8 Flash Cyber is positioned as Google's "most capable cybersecurity model with frontier-level performance in vulnerability detection and automated patching." This model is currently available only to trusted defenders participating in Google's program. The safety features of 3.8 Flash include safeguards against chemical, biological, radiological, nuclear, and cyber offenses. Additionally, 3.8 Flash Cyber has a more permissive set of mitigations tailored to the needs of cybersecurity professionals.

Google emphasizes that both models, despite their distinct deployment environments, share the same foundational intelligence and are accelerated by long-running agentic loops designed to continuously evaluate and refine the models. The significant improvements in coding and reasoning capabilities are attributed to innovations such as rigorous training in the challenging domain of cybersecurity.