Google has unveiled its latest models under the Gemini 3.8 framework, touted as the company’s most advanced coding and logical reasoning models to date. This announcement comes just three weeks after the introduction of version 3.7 Flash, marking the third major launch in a rapid deployment cycle over six weeks.
The new Gemini 3.8 models maintain the high processing speeds and cost efficiency of their predecessor while offering two distinct editions: a general-purpose workhorse model designed for a wide range of applications and a specialized model focused on cybersecurity. This strategic expansion reflects Google’s commitment to advancing AI capabilities in both software development and security sectors.
“Today we’re introducing a new 3.8 Flash model, our 3rd Flash release in just 6 wks. It delivers significant leaps from 3.7 Flash across software engineering, agentic tasks, and multi-step reasoning. On DeepSWE v1.1, it outperforms most larger frontier models in autonomously solving complex engineering problems end to end, at a fraction of the cost,” said CEO Sundar Pichai in a post on X.
Both these models share a common intelligent foundation strengthened by iterative training in cybersecurity, enhanced by recursive agentic loops that continually assess and sharpen performance, according to Google.
Designed for complex, multi-stage reasoning, software development, and autonomous agent workflows, the standard Gemini 3.8 Flash delivers a major step up from version 3.7 while keeping initial pricing fixed at $0.75 per million input tokens and $3.75 per million output tokens, Google says. As per the company, Gemini 3.8 Flash managed to surpass most larger frontier AI systems at a much lower cost on the DeepSWE v1.1 evaluation test, which measures end-to-end software engineering capabilities over extended tasks. The second variant, Gemini 3.8 Flash Cyber, is tailored directly for defensive digital security teams and is being distributed to verified defenders through Google’s newly unveiled Fairwind Program.
Google emphasised that development for the Cyber model focused heavily on vulnerability remediation rather than offensive exploitation techniques.
“We’re also introducing Gemini 3.8 Flash Cyber, our most capable cybersecurity model. That includes achieving 86.2% on the important CyberGym industry benchmark, plus 47.2% on CWE-Bench for patching. We saw a 70%+ success rate in discovering vulnerabilities across 20 programming languages on our internal benchmark,” Pichai said in a separate post. On the same benchmarks, Anthropic’s Opus 5 and GPT-5.6 Sol scored 83.8% and 83.6%, respectively. Google says it has already integrated Gemini 3.8 Flash Cyber into active production defenses. Testing by Google’s Chrome Security division revealed that the model generated 2.6 times more accurate vulnerability patches for the Chrome browser than the top commercial models of much larger scale. For example, in evaluations on real Chrome security bugs, 3.8 Flash Cyber produced 2.6x more correct patches for vulnerabilities than larger models.
Because of its advanced capabilities, we’re giving Government agencies & cybersecurity partners access through the Fairwind program,” Pichai added. It shows frontier-level performance in discovering vulnerabilities and patching them at scale, with Flash-level speed & pricing. “Beyond benchmarks, we’re seeing impressive real-world use cases. Get the latest technology news and updates. Download the TOI App.