Gemini 3.8 Flash Brings Frontier-Level Performance at Flash Prices

Google has introduced Gemini 3.8 Flash and Gemini 3.8 Flash Cyber, two new versions of its Flash model designed for advanced reasoning, coding, autonomous agents, and cybersecurity.

Key points

• The model outperforms Gemini 3.7 Flash and other frontier models on benchmarks including Vals Finance Agent V2 and Harvey's Legal Agent Benchmark.
• Consumers can employ Gemini 3.8 Flash through the Gemini app, AI Mode in Google Search, and Gemini in Google Sheets with Google AI Pro and Ultra subscriptions.
• As well as Google Antigravity and Android Studio, developers can employ Gemini 3.8 Flash through the Gemini API in Google AI Studio.
• While Gemini 3.8 Flash Cyber remains restricted to trusted cybersecurity defenders, gemini 3.8 Flash is broadly available throughout Google's developer, enterprise, and consumer products.
• While Flash Cyber is designed specifically for cybersecurity defenders, gemini 3.8 Flash is positioned as Google's most capable workhorse model in the Flash series.

Google notes it can approach the performance of more expensive frontier models while retaining the speed and lower cost associated with the Flash series.

While Flash Cyber is designed specifically for cybersecurity defenders, gemini 3.8 Flash is positioned as Google's most capable workhorse model in the Flash series. Both models employ the same underlying intelligence and are built to perform further reasoning and iterative tool calls on complex tasks. Google Launches Pics as Canva Rival But With AI Image Editing. Gemini 3.8 Flash.

Gemini 3.8 Flash brings significant improvements over Gemini 3.7 Flash in software engineering, agentic tasks, and multi-step reasoning.

Notably, the model is designed for long-horizon coding and autonomous agents. On the DeepSWE v1.1 software engineering benchmark, Google notes Gemini 3.8 Flash outperforms most larger frontier models at solving complex engineering difficulties end to end.

It additionally improves performance in specialized professional fields. The model outperforms Gemini 3.7 Flash and other frontier models on benchmarks including Vals Finance Agent V2 and Harvey's Legal Agent Benchmark. It scores 54.9% on HLE-Verified. It tests multi-step reasoning throughout STEM, humanities, and professional subjects.

Meanwhile, the improvements come partly from giving the model more room to reason. On complex tasks, Gemini 3.8 Flash can perform extra reasoning steps and repeatedly call tools to evaluate and refine its work. While developers can employ lower effort levels when compute efficiency is more notable, higher effort levels can therefore consume more tokens.

Google additionally demonstrated the model building games, interactive 3D visualizations, a playable DOS version of Google Maps, and topographic maps using real datasets.

Gemini 3.8 Flash is available at the same introductory cost as Gemini 3.7 Flash.

As well as Google Antigravity and Android Studio, developers can employ Gemini 3.8 Flash through the Gemini API in Google AI Studio. It is additionally available for UI generation through Stitch. Enterprise customers can access it through Gemini Enterprise.

Consumers can employ Gemini 3.8 Flash through the Gemini app, AI Mode in Google Search, and Gemini in Google Sheets with Google AI Pro and Ultra subscriptions. Claude Fable 5.1 Brings Massive Performance Gains at Lower Cost. Gemini 3.8 Flash Cyber.

Gemini 3.8 Flash Cyber is a specialized version built for cybersecurity defenders. Google describes it as its most capable cybersecurity model, with frontier-level performance in vulnerability discovery and automated patching.

Unlike the standard Flash model, Flash Cyber is available only to trusted defenders through Google's Fairwind Program since it has more permissive cybersecurity safeguards.

Gemini 3.8 Flash Cyber achieves frontier-level results on CyberGym, a benchmark for autonomous vulnerability discovery. Google notes it surpasses Gemini 3.5 Flash Cyber as well as significantly larger frontier models.

On Google's internal benchmark covering complex codebases throughout 20 programming languages, the model achieves a vulnerability-discovery success rate above 70%.

In practice, the model additionally focuses on fixing vulnerabilities. While costing significantly less per rollout, on the CWE-Bench patching benchmark, it records a 47.2% Pass@1 score, close to a leading frontier model's 47.8%.

Google notes its own teams are already using the model to secure software. While Google's Cloud Vulnerability Research team employed it to identify a critical foundational vulnerability in less than two hours, the Chrome Security team discovered that it produced 2.6 times more correct patches for Chrome vulnerabilities than the best commercial models tested.

Google has not provided a separate public cost for Gemini 3.8 Flash Cyber. Access is instead being provided to selected trusted defenders through the Fairwind Program.

Gemini 3.8 Flash Cyber is not generally available to consumers or developers. Google is providing prioritized access through the Fairwind Program to trusted administration authorities, critical infrastructure operators, and software maintainers. Eligible organizations can apply for access through the program.

Both Gemini 3.8 models include stronger protection against prompt-injection attacks. Google notes the models additionally include safeguards addressing cyber misuse and chemical, biological, radiological, and nuclear risks.

While Gemini 3.8 Flash Cyber remains restricted to trusted cybersecurity defenders, gemini 3.8 Flash is broadly available throughout Google's developer, enterprise, and consumer products. Stay Connected with ProPakistani.

Obtain the latest tech news, telecom insights, and product launches wherever you prefer. Follow on Google Discover.

Add as a preferredSource on Google Follow on Google News Join WhatsApp.

Add ProPakistani to Preferred Sources and see more of our stories in Google Search and Top Stories.

For now, gemini 3.8 Flash Brings Frontier-Level Performance at Flash Prices remains the part of the story worth watching, and further updates are likely as more details are confirmed.