Google Introduces Gemini 3.8 Flash and 3.8 Flash Cyber With Stronger AI and Security Skills

Google has introduced two new AI models, Gemini 3.8 Flash and Gemini 3.8 Flash Cyber. The company says the new models bring better reasoning, coding, and agentic capabilities while keeping the speed and lower cost associated with the Flash series.

Gemini 3.8 Flash is designed as a general-purpose model for developers, businesses and consumers. Google says it delivers major improvements over Gemini 3.7 Flash, particularly when handling complex software engineering tasks, autonomous agents and multi-step reasoning.

Despite these improvements, Google is keeping the introductory pricing at the same level as Gemini 3.7 Flash. The model costs $0.75 per million input tokens and $3.75 per million output tokens.

Google Introduces Gemini 3.8 Flash and 3.8 Flash Cyber With Stronger AI and Security Skills

A major focus of Gemini 3.8 Flash is long-running coding work. Google says the model can independently work through complex software engineering problems and complete tasks from start to finish. On the DeepSWE v1.1 benchmark, the company says Gemini 3.8 Flash can outperform several larger and more expensive frontier models.

The model also targets professional workloads that require detailed analysis. Google reports strong results on finance and legal agent benchmarks, along with a 54.9% score on HLE-Verified. This shows the model’s ability to work through difficult questions across areas such as science, humanities, and professional fields.

Google says Gemini 3.8 Flash achieves these results by spending more effort on complicated tasks. It can perform additional reasoning steps and repeatedly use tools when needed. This approach can improve results, although it may also increase token usage.

Developers who want to reduce computing costs can select lower effort levels. Google will also continue supporting Gemini 3.7 Flash for workloads where efficiency is more important than maximum performance.

The second model, Gemini 3.8 Flash Cyber, focuses specifically on cybersecurity. Google describes it as its most capable cybersecurity model yet, with a focus on finding vulnerabilities and automatically fixing security problems.

The model has demonstrated strong results on CyberGym, a benchmark designed to test autonomous vulnerability discovery. Google says Gemini 3.8 Flash Cyber outperformed its previous 3.5 Flash Cyber model as well as several much larger frontier models.

Google also tested the model on an internal benchmark covering complex software projects written in 20 programming languages. The company says the model achieved a success rate of more than 70% when identifying vulnerabilities across these codebases.

Automated vulnerability repair is another key feature. On the CWE-Bench benchmark, Gemini 3.8 Flash Cyber recorded a 47.2% pass@1 score. Google says this is close to the 47.8% result achieved by a leading frontier model, while Gemini 3.8 Flash Cyber costs significantly less to operate.

Google is already using the cybersecurity model internally. The company says its Chrome Security team found that Gemini 3.8 Flash Cyber generated 2.6 times more correct vulnerability patches than the best commercial models tested, despite those models being much larger.

Google also says its Cloud Vulnerability Research team used the model to identify a critical vulnerability in less than two hours. According to the company, research of this type can normally take months.

See Also: Google Gemini May Let Users Turn Off Visible Watermarks on AI-Generated Content

Security remains a major part of the new models. Gemini 3.8 Flash includes safeguards designed to limit harmful use involving cyber offense and CBRN-related areas. Gemini 3.8 Flash Cyber has more permissive cybersecurity protections because it is intended for trusted defenders.

For this reason, Google is making Gemini 3.8 Flash Cyber available through its Fairwind Program to selected government authorities, critical infrastructure operators, and software maintainers.

Gemini 3.8 Flash is already available through several Google products and developer platforms. Developers can access it through the Gemini API, Google AI Studio, Android Studio and Google Antigravity. Enterprises can use it through Gemini Enterprise, while eligible consumers can access it through the Gemini app, Google Search’s AI Mode, and Gemini in Google Sheets.

With the two new models, Google is expanding Gemini’s focus beyond traditional chatbot use. Gemini 3.8 Flash is aimed at demanding coding, reasoning, and agent-based tasks, while the Cyber version targets organizations that need advanced tools for discovering and fixing software vulnerabilities.