# Google Introduces Gemini 3.8 Flash and 3.8 Flash Cyber Next-gen AI models revolutionize cybersecurity and software engineering with unprecedented performance. - **Source:** [Hacker News (news.ycombinator.com)](https://brifing.si/source/13c923cd-6d99-442b-9b8b-1181f4d8cd91) - **Published:** 2026-09-02 - **Topic:** [Data & AI](https://brifing.si/c/data-ai) - **Original article:** - **On Brifing:** https://brifing.si/s/109a33a6-8d9e-4988-9940-c94d71775808 ## Introduction of Gemini 3.8 Flash Google announced Gemini 3.8 Flash, building upon the 3.7 Flash. This new model offers significant coding and reasoning improvements over its predecessor. Intensive training in the cybersecurity domain particularly boosted the model's performance. While approaching the performance of higher-cost frontier models, 3.8 Flash maintains the same speed and low-cost advantage. In challenging software engineering tests like DeepSWE v1.1, it outperforms larger models in end-to-end complex problem solving. ## Performance Boost and Innovations Gemini 3.8 Flash demonstrates the reliability required for critical enterprise autonomy across specialized knowledge domains. It surpasses 3.7 Flash and other frontier models in benchmarks like Vals Finance Agent V2 and Harvey's Legal Agent Benchmark. Achieving a 54.9% success rate on HLE-Verified, it proves its multi-step reasoning capability across STEM, humanities, and professional fields. These advancements are enabled by the model executing extra reasoning steps and using tools iteratively on complex tasks. ## Computational Efficiency and Flexibility Gemini 3.8 Flash can sometimes use more tokens to maximize performance. However, for applications where computational efficiency is priority, developers can utilize lower effort levels to minimize token overhead or continue relying on Gemini 3.7 Flash, which remains fully supported for efficiency-first workloads. This flexibility facilitates adaptation to different use cases. ## Creative Applications with Google Antigravity Gemini 3.8 Flash created a game using a looping instruction in Google Antigravity. Using puzzles, environmental storytelling, and textures generated with Nano Banana, it crafted an immersive 3D level where players navigate a castle as a wizard. It also built a fully functional DOS version of Google Maps with a single prompt, complete with locations, directions, and Street View functionality. ## Scientific Visualization and Hardware Analysis Gemini 3.8 Flash generated topographic maps of famous geographical sites using real datasets from the U.S. Geological Survey, offering real-time cross-sections, 2D projections, and scientific explanations. Hardware Anatomy, built in Google AI Studio, visualizes physically-proportioned teardowns of hardware devices with realistic Three.js renderings. Devices can be decomposed into layers for inspection using a deconstruction slider. ## Advantages of Gemini 3.8 Flash Cyber Available to trusted defenders via the Fairwind Program, Gemini 3.8 Flash Cyber demonstrates frontier-level performance in autonomous vulnerability discovery on the CyberGym benchmark. It outperforms both 3.5 Flash Cyber and significantly larger frontier models. When tested against a comprehensive internal benchmark spanning 20 programming languages, it shows an impressive leap over previous models, achieving over 70% success rate in discovering diverse vulnerabilities. ## Defense-Focused Approach Gemini 3.8 Flash Cyber equips defenders with expert capabilities that provide an advantage over attackers. On CWE-Bench, run by Collinear, it sits on the Pareto frontier for patching capabilities. With a pass@1 of 47.2%, it performs comparably to a leading frontier model at 47.8%, while being offered at significantly lower cost. Google has already begun using this model to secure its own code. ## Security and Pricing Gemini 3.8 Flash includes safeguards against misuse in Chemical, Biological, Radiological, and Nuclear (CBRN) and cyber offense domains, enabling beneficial use cases per the Frontier Safety Framework. 3.8 Flash Cyber comes with a more permissive set of mitigations for cybersecurity and is only available to trusted defenders. It also shows significant improvement in prompt injection robustness as measured by Gray Swan. The introductory pricing expires December 31, 2026. ## Sources covering this story - https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._