Google Launches Gemini 3.8 Flash Alongside New Cybersecurity Model

Google has launched Gemini 3.8 Flash, a new AI model for software engineering, agentic tasks and multi-step reasoning, alongside Gemini 3.8 Flash Cyber, a version designed for cybersecurity applications.

Gemini 3.8 Flash is available at the same introductory price as Gemini 3.7 Flash — $0.75 per million input tokens and $3.75 per million output tokens. Gemini 3.8 Flash Cyber is being offered to trusted defenders through Google’s Fairwind Program.

Gemini 3.8 Flash is available to developers through the Gemini API, Google AI Studio and Android Studio, and to enterprises through Gemini Enterprise. Consumers can access it through the Gemini app, AI Mode in Google Search and Gemini in Google Sheets.

Google said Gemini 3.8 Flash is its “best reasoning & coding model yet” while maintaining the speed and low cost of its predecessor. The company said the model delivers improvements in software engineering, agentic workflows and reasoning across specialised domains.

On the DeepSWE v1.1 benchmark for long-horizon software engineering, Google said Gemini 3.8 Flash outperformed most larger frontier models in autonomously solving complex engineering problems end to end.

The model’s performance comes partly from using additional reasoning steps and iterative tool calls on complex tasks. Google said Gemini 3.8 Flash “works harder” on such tasks and may use more tokens at higher effort levels. Developers can select lower effort levels when compute efficiency is a priority.

Gemini 3.8 Flash Cyber is focused on vulnerability discovery and automated patching. On CyberGym, a benchmark for autonomous vulnerability discovery, Google said the model achieved frontier-level performance and surpassed Gemini 3.5 Flash Cyber as well as larger frontier models.

In an internal benchmark covering vulnerabilities across codebases written in 20 programming languages, Gemini 3.8 Flash Cyber achieved a success rate of more than 70%, according to Google.

For automated patching, the model recorded a pass@1 score of 47.2% on CWE-Bench, compared with 47.8% for a leading frontier model, while Google said it was available at a lower cost.

Google said it is already using Gemini 3.8 Flash Cyber internally. Its Chrome Security team found that the model produced 2.6 times more correct patches for Chrome vulnerabilities than the best commercial models tested, while Google's Cloud Vulnerability Research team used it to identify a critical foundational vulnerability in less than two hours.

The cybersecurity model is not being made generally available. Google said Gemini 3.8 Flash Cyber will be provided through the Fairwind Program to trusted government authorities, critical infrastructure operators and software maintainers.

Google has also added safeguards against misuse involving cyber offence and chemical, biological, radiological and nuclear domains. The company said Gemini 3.8 models have improved resistance to prompt-injection attacks.