Google has released Gemini 3.8 Flash, and the interesting part is not the model itself but the calendar. It arrived on 2 September 2026, three weeks after Gemini 3.7 Flash and the company’s third Flash release in six weeks. Anyone building products on these tools is now working on ground that shifts roughly every fortnight.
Google calls it its “most intelligent workhorse model” and says it improves on software engineering, agentic tasks and multi-step reasoning in specialised fields. On the DeepSWE v1.1 benchmark for long-horizon software engineering, the company says it outperforms most larger frontier models at solving complex engineering problems from start to finish. Google also reports gains in professional and quantitative work such as finance and legal tasks, and says the model often comes close to pricier frontier systems while keeping the speed and low cost of the Flash line.
Working harder, for the same price
The company is candid about where the gains come from. “These performance gains stem from a core design choice: 3.8 Flash works harder,” Google said. On complex tasks the model takes extra reasoning steps and calls tools repeatedly, and at higher effort levels it may use more tokens to get the best result. The introductory price is unchanged from 3.7 Flash, at $0.75 per million input tokens and $3.75 per million output tokens. Developers can reach it through the Gemini API, Google AI Studio, Android Studio and other Google platforms.
A model for defenders
Alongside it came Gemini 3.8 Flash Cyber, built to help security teams find and fix software vulnerabilities. Google describes it as its most capable cybersecurity model yet and says it beat both the earlier Gemini 3.5 Flash Cyber and considerably larger frontier models on CyberGym, a test of autonomous vulnerability discovery.
Unlike the standard model, the Cyber version is not open to everyone at first. It goes initially to trusted defenders through Google’s Fairwind Program, among them government authorities, critical infrastructure operators and software maintainers. Google says both models carry safeguards against misuse involving chemical, biological, radiological and nuclear threats, as well as offensive cyber activity.
