Google has launched Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, its latest models for coding, agents, and cybersecurity. The company announced both models on September 2, 2026, as the release comes only three weeks after the launch of Gemini 3.7 Flash. Google mentions in the announcement, “Our newest Gemini models deliver next-generation intelligence for agentic workflows and cybersecurity.”
According to the company, Gemini 3.8 Flash is its most intelligent workhorse model to date that targets software engineering, autonomous agents, and complex multi-step reasoning. On the other hand, Gemini 3.8 Flash Cyber focuses specifically on cybersecurity, focusing on discovering vulnerabilities and automating software patching. Here is how Gemini 3.8 Flash performs across benchmarks:
The release of Gemini 3.8 Flash and 3.8 Flash Cyber also highlights Google’s third Flash model launch within six weeks. The quick advancements indicate the company’s push toward faster AI development.
Gemini 3.8 Flash Targets Coding and AI Agents
Gemini 3.8 Flash builds on the capabilities introduced with Gemini 3.7 Flash. Google said the new model delivers substantial improvements across several demanding workflows, including software engineering, agentic tasks, and specialized knowledge work. Gemini 3.8 Flash can take additional reasoning steps on difficult problems. It can also call tools repeatedly when completing complex workflows alongside managing long-horizon coding tasks.
This approach can improve performance on tasks requiring several connected actions. However, it can also increase token usage. Google therefore allows developers to use lower effort levels when computing efficiency matters. Developers can also continue using Gemini 3.7 Flash for efficiency-focused workloads.
The company reported strong results across several benchmarks. Gemini 3.8 Flash scored 54.9% on HLE-Verified. It also outperformed Gemini 3.7 Flash on Vals Finance Agent V2. The model also showed stronger results on Harvey’s Legal Agent Benchmark. On DeepSWE v1.1, Google reported strong performance on long-horizon software engineering tasks. The benchmark evaluates an AI model’s ability to solve complex engineering problems autonomously.
Gemini 3.8 Flash Cyber Focuses on Vulnerability Detection
Google also introduced Gemini 3.8 Flash Cyber for defensive cybersecurity operations. The specialized model focuses on vulnerability discovery and automated patching. Google said it prioritizes defensive capabilities rather than offensive exploitation while demonstrating frontier-level performance on the CyberGym vulnerability discovery benchmark. It also exceeded previous Gemini cybersecurity models on Google’s internal evaluations.
Google tested the model across complex codebases covering 20 programming languages. The company reported a vulnerability discovery success rate above 70%. For automated patching, Gemini 3.8 Flash Cyber achieved a 47.2% pass@1 score on CWE-Bench. Google said this result places the model on the benchmark’s Pareto frontier.
Google is already applying the model to its own security work. The Chrome Security team reported 2.6 times more correct patches than larger commercial models. Google also shared that its Cloud Vulnerability Research team found a critical vulnerability within two hours with the help of 3.8 Flash Cyber. Such research can normally require months of investigation.
Access Comes with Different Limits in Gemini 3.8 Flash and 3.8 Flash Cyber
Gemini 3.8 Flash is available to developers and enterprise customers. Consumers can access it through Google AI Pro and Ultra subscriptions. Developers can use the model through Google AI Studio, the Gemini API, Android Studio, and Google Antigravity. Enterprises can access it through Gemini Enterprise.
Gemini 3.8 Flash Cyber has more restricted availability. Google is offering it through the new Fairwind Program. The program provides prioritized access to trusted government authorities, critical infrastructure operators, and software maintainers.
Competitive Pricing of the Flash Models is a Major Advantage:
Gemini 3.8 Flash carries the same introductory pricing as Gemini 3.7 Flash. Google is charging $0.75 per million input tokens and $3.75 per million output tokens. The introductory pricing is available through December 31, 2026. The model’s pricing is part of Google’s broader effort to make advanced AI practical for large-scale workloads. However, higher reasoning efforts can increase total token consumption.
The release of Gemini 3.8 Flash and 3.8 Flash Cyber shows Google targeting both everyday AI workloads and specialized enterprise security operations. Stay aligned with the latest AI breakthroughs and updates with KnowledgeNile.
Recommended For You:
Google Releases Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber to Power Next-Gen AI Agents
