On September 2, 2026, Google unveiled its latest AI model, Gemini 3.8 Flash, following the quick succession of its predecessors Gemini 3.6 Flash and 3.7 Flash. This new iteration emphasizes next-generation intelligence, particularly in the realms of cybersecurity and agentic workflows.
Overview of Gemini 3.8 Flash
Gemini 3.8 Flash is designed to be Google’s most advanced workhorse model to date. It has been tailored for developers who are creating agentic applications, providing them with enhanced capabilities. Additionally, Google AI Pro and Ultra subscribers can access this model through the Gemini app.
Performance Metrics
According to Google, Gemini 3.8 Flash has shown superior performance on the DeepSWE v1.1 benchmark, where it achieved a score of 73.7%. This places it just below Claude Opus 5, which scored 74%, and significantly ahead of previous models like Gemini 3.7 Flash and GPT-5.6 varieties.
The model also excelled in various other benchmarks, including financial analysis, legal workflows, information synthesis, and video comprehension tasks. Google attributes these performance gains to the model's improved reasoning capacity, enabling it to tackle complex problems more diligently.
Pricing Structure
Gemini 3.8 Flash is priced at an introductory rate of $0.75 for every million input tokens and $3.75 for output tokens. However, these prices will adjust to $1.50 and $7.50 after the introductory period, respectively. Despite this increase, Gemini 3.8 Flash remains competitively priced compared to models from other developers.
Introducing Gemini 3.8 Flash Cyber
In addition to the standard model, Google has launched Gemini 3.8 Flash Cyber, which is specifically designed for cybersecurity applications. However, access to this version is restricted to members of the Fairwind Program, which includes trusted government entities and critical infrastructure operators.
Gemini 3.8 Flash Cyber scored an impressive 86.2% on industry-standard benchmarks for identifying vulnerabilities in C/C++ programming. This score surpasses both its predecessor and larger models like Mythos 5. Furthermore, it demonstrated an impressive 71% success rate in real-world vulnerability discovery across multiple programming languages, significantly higher than the previous models.
Automated Patching Capabilities
One of the standout features of Gemini 3.8 Flash Cyber is its automated patching functionality. This allows users to generate effective patches for vulnerabilities, with the model reportedly producing 2.6 times more accurate patches for Chrome vulnerabilities than its larger commercial counterparts.
While most everyday users may not interact directly with Gemini 3.8 Flash, its advancements can help improve the overall safety and reliability of technology. The model is currently rolling out to developers, enterprise clients, and Google AI Pro and Ultra subscribers, while Gemini 3.8 Flash Cyber is available exclusively to Fairwind Program participants.
Source: https://www.androidauthority.com/gemini-3-8-flash-google-ai-model-3706483/
