Google DeepMind
Build with Gemini
Try Gemini

Gemini 3.8 Flash Cyber

Finding and fixing vulnerabilities quickly and efficiently

Learn more

Gemini 3.8 Flash Cyber: our most capable cybersecurity model, with frontier-level performance in vulnerability detection, and automated patching.

Capabilities

Gemini 3.8 Flash Cyber provides a decisive advantage in today’s complex cybersecurity landscape, with the Flash speed and cost that enables quick iteration.

The model is designed specifically for defenders, has frontier-level performance in vulnerability detection, and automated patching.

A minimalist list search icon in a white rounded square against a soft blue gradient background.Minimalist search list icon inside a black rounded square on a dark blue gradient background.

Autonomous vulnerability discovery

Demonstrates frontier-level performance in autonomous vulnerability discovery, with the Flash speed and cost that enables quick iteration. Capable of navigating complex codebases spanning 20 programming languages to uncover critical hidden vulnerabilities.

Minimalist icon of crossed adhesive bandages on a white rounded square, centered on a soft blue gradient background.An icon of crossed bandages inside a dark rounded square, centered on a dark blue gradient background.

Automated patching

Generates validated, high-quality code fixes automatically after discovering vulnerabilities. So defenders can neutralize potential threats before attackers discover and exploit them.

Fairwind Program

The Fairwind Program gives trusted partners a critical head start against AI-driven cyber threats, while making sure models are deployed responsibly.

Learn more

Performance

Gemini 3.8 Flash Cyber, available to a set of trusted defenders via the Fairwind Program, provides a decisive advantage in today’s complex cybersecurity landscape, with the Flash speed and cost that enables quick iteration.

A bar chart comparing performance on the CyberGym Pass@1 vulnerability discovery benchmark. Gemini 3.8 Flash Cyber leads with 86.2%, followed by GPT-5.5-Cyber (85.6%), Mythos 5 (83.8%), GPT-5.6 Sol (83.6%), and Gemini 3.5 Flash Cyber (77.5%).A bar chart comparing performance on the CyberGym Pass@1 vulnerability discovery benchmark. Gemini 3.8 Flash Cyber leads with 86.2%, followed by GPT-5.5-Cyber (85.6%), Mythos 5 (83.8%), GPT-5.6 Sol (83.6%), and Gemini 3.5 Flash Cyber (77.5%).

CyberGym Pass@1

On the standard industry benchmark for finding vulnerabilities, CyberGym, Gemini 3.8 Flash Cyber demonstrates frontier-level performance in autonomous vulnerability discovery. It surpasses both 3.5 Flash Cyber as well as significantly larger frontier models.

A bar chart showing real-world vulnerability discovery performance (Pass@1 across 20 languages) for three models: Gemini 3.8 Flash Cyber (71.0%), Gemini 3.7 Flash (58.9%), and Gemini 3.5 Flash Cyber (46.6%).A bar chart comparing Gemini model performance on an internal vulnerability discovery benchmark across 20 languages: Gemini 3.8 Flash Cyber (71.0%), Gemini 3.7 Flash (58.9%), and Gemini 3.5 Flash Cyber (46.6%).

Real-world vulnerability discovery

We also evaluated Gemini 3.8 Flash Cyber against a comprehensive internal benchmark in which the model has to discover a wide range of vulnerabilities across complex codebases spanning 20 programming languages. Here, the model showcases an impressive leap over our previous models and reaches a success rate exceeding 70%.

Scatter plot comparing AI models on "CWE-Bench Pass@1 vs. Cost per Rollout." A blue Pareto curve highlights top performers like Gemini 3.8 Flash Cyber, which offers high accuracy (around 47%) at a lower cost (~$3.60) compared to competitors like Fable 5.Scatter plot comparing AI models on "CWE-Bench Pass@1 vs. Cost per Rollout." A blue Pareto curve highlights top performers like Gemini 3.8 Flash Cyber, which offers high accuracy (around 47%) at a lower cost (~$3.60) compared to competitors like Fable 5.

CWE-Bench Pass@1 vs. Cost per Rollout

CWE-Bench, run by Collinear, is a challenging external benchmark for patching capabilities. On this benchmark, Gemini 3.8 Flash Cyber is on the Pareto frontier: with a pass@1 of 47.2% compared to Fable 5 at 47.8%, yet offered at a significantly lower cost.

Stacked bar chart showing prompt injection attack success rates on the Gray Swan IPI benchmark. Gemini 3.8 Flash Cyber ranks among the most robust models with a low 6.0% success rate (lower is better).Stacked bar chart showing prompt injection attack success rates on the Gray Swan IPI benchmark. Gemini 3.8 Flash Cyber ranks among the most robust models with a low 6.0% success rate (lower is better).

Gray Swan IPI benchmark

Gemini 3.8 models have also made a significant leap in prompt injection robustness as measured by Gray Swan, protecting Gemini model users from prompt-injection related malicious attacks.