Google Unveils Gemini 3.8 Flash, Bringing Faster AI Reasoning and a Dedicated Cybersecurity Model
Google has expanded its Gemini lineup once again with the launch of Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, introducing a new generation of models focused on advanced reasoning, software development, autonomous AI agents and cybersecurity.
The announcement comes just three weeks after Gemini 3.7 Flash, making Gemini 3.8 the third Flash release from Google in only six weeks. Google describes Gemini 3.8 Flash as its most capable Flash model for reasoning and coding so far, while maintaining the speed and relatively low cost associated with the Flash family.
Gemini 3.8 Flash Is Designed to “Work Harder”
The standard Gemini 3.8 Flash is aimed at developers, businesses and users who need an AI model capable of handling complicated tasks without the cost and latency traditionally associated with larger frontier models.
Google says the new model delivers major improvements over Gemini 3.7 Flash in software engineering, agentic workflows and multi-step reasoning. One of its biggest changes is the model’s ability to spend additional computational effort on difficult problems, effectively allowing it to reason through a task more extensively when necessary.
That approach is particularly important for AI agents. Instead of simply answering a prompt and stopping, Gemini 3.8 Flash can perform longer sequences of reasoning and tool calls, allowing it to tackle tasks that require multiple steps.
Google demonstrated the model creating functional software and interactive experiences, including a playable 3D game and a working DOS-style version of Google Maps using a single prompt through its development tools.
A Major Push Into Coding
Coding is one of the areas where Google is putting the biggest emphasis with Gemini 3.8 Flash.
The company says the model performs strongly on DeepSWE v1.1, a benchmark designed around long-horizon software engineering tasks. According to Google, Gemini 3.8 Flash can outperform many larger frontier models on complex engineering problems while operating at a substantially lower cost.
The model is also being positioned for professional workloads that require deeper analysis. Google reports improvements on benchmarks covering areas such as financial analysis, legal reasoning and broader multi-step knowledge tasks.
For developers, that could make Gemini 3.8 Flash particularly useful for building AI coding agents, automated development workflows and applications that can independently complete longer sequences of tasks.
Gemini 3.8 Flash Cyber Takes Aim at Software Vulnerabilities
Alongside the general-purpose model, Google has introduced Gemini 3.8 Flash Cyber, a specialized version designed specifically for cybersecurity.
Rather than being marketed as a general consumer AI assistant, Flash Cyber is focused on helping security teams discover vulnerabilities and develop fixes. Google says the model achieved frontier-level performance on vulnerability discovery and demonstrated significant improvements in automated patching.
Google’s own security teams are already using the technology. The company says its Chrome Security team found that Gemini 3.8 Flash Cyber generated 2.6 times more correct patches for Chrome vulnerabilities than the best commercial models tested by the team. Google also says its Cloud Vulnerability Research team used the model to identify a critical foundational vulnerability in less than two hours — work that traditionally could take months.
The model is currently being distributed through Google’s Fairwind Program, which provides access to trusted defenders such as government authorities, critical infrastructure operators and software maintainers.
Google Keeps the Flash Pricing Strategy
Despite the performance improvements, Google isn’t dramatically increasing the introductory price.
Gemini 3.8 Flash is launching at $0.75 per million input tokens and $3.75 per million output tokens, matching the introductory pricing of Gemini 3.7 Flash. Google notes that these prices are promotional through December 31, 2026, after which they are scheduled to increase to $1.50 per million input tokens and $7.50 per million output tokens.
There is, however, an important consideration: because Gemini 3.8 Flash may use more tokens and perform additional reasoning steps on difficult tasks, the actual cost of completing a task can be higher even if the per-token price remains unchanged.
Where Can You Use Gemini 3.8 Flash?
Google is making the standard model available across much of its AI ecosystem.
Developers can access Gemini 3.8 Flash through the Gemini API, Google AI Studio, Android Studio and Google Antigravity, while enterprise users can access it through Gemini Enterprise. Consumers with Google AI Pro and Ultra subscriptions can also use the model through the Gemini app, AI Mode in Google Search and Gemini in Google Sheets.
Flash Cyber, meanwhile, remains considerably more restricted because of the sensitive nature of its cybersecurity capabilities.
Google Is Accelerating Its Gemini Flash Releases
Perhaps the most interesting part of this announcement is how quickly Google is iterating.
Gemini 3.8 Flash arrives only weeks after 3.7 Flash, continuing Google’s increasingly rapid rollout of specialized Flash models. The company appears to be betting that smaller, faster models with strong reasoning capabilities can handle a growing share of workloads that once required much larger systems.
For users, that could mean increasingly capable AI without having to sacrifice speed. For developers, it could mean cheaper and more practical AI agents capable of completing increasingly complicated tasks on their own.
And with Gemini 3.8 Flash now joining Google’s lineup, the race toward the next generation of Gemini models has become even more interesting.
Gemini 3.8 Flash isn’t simply another incremental update. Google is positioning it as a faster, more capable workhorse for coding, reasoning and AI agents — while Flash Cyber represents a more specialized attempt to put advanced AI directly in the hands of cybersecurity defenders.
