Gemini 3.8 Operational

Gemini 3.8 is Google's reasoning and coding model, available in Flash and Flash Cyber versions. Gemini 3.8 Flash significantly improves software engineering, agent tasks, and multi-step reasoning while maintaining the same speed and low price as its predecessor. It approaches or even surpasses larger frontier models in benchmarks covering coding, finance, law, and other areas. Gemini 3.8 Flash Cyber is designed specifically for cybersecurity, offering leading vulnerability discovery and automated patch remediation capabilities. It is available only to certified security defenders through the Fairwind Program.

Long-horizon software engineering: Autonomously completes complex engineering problems end to end, approaching or even surpassing larger frontier models in benchmarks such as DeepSWE.
Autonomous agent tasks: Supports complex, multi-step agent workflows with iterative tool calls and reasoning, suitable for terminal coding, computer operation, and other scenarios.
Professional-domain reasoning: Demonstrates enterprise-grade reliability in financial analysis, legal workflows, and cross-disciplinary expert reasoning.
Cybersecurity offense and defense: The Flash Cyber version focuses on autonomous vulnerability discovery and automated patch remediation, reaching frontier-level performance on industry benchmarks such as CyberGym and CWE-Bench.
Native multimodality: Supports text, code, image, video, and other modalities as input, and can generate rich-media content such as 3D visualizations, games, and interactive applications.

Long-running agentic loops: Gemini 3.8 introduces long-running agentic loops, enabling the model to recursively evaluate and refine its own output during task execution. Unlike a single forward pass, the model continues iterating on complex tasks by performing additional reasoning steps, repeatedly calling external tools, and dynamically adjusting its strategy based on intermediate results. This enables deeper execution and greater accuracy in multi-step reasoning and code-generation tasks.
Dynamic compute allocation and reasoning-effort adjustment: The model follows a design philosophy of allocating more compute when needed, demonstrating greater diligence on complex tasks. It automatically assigns more reasoning resources based on task difficulty, potentially consuming more Tokens at higher effort levels to maximize performance. Developers can also select lower reasoning-effort levels to reduce Token usage, flexibly balancing performance and cost.
Specialized cybersecurity training: Gemini 3.8 Flash and Flash Cyber share the same underlying foundational intelligence. Their significant improvements in coding and reasoning partly stem from rigorous training in cybersecurity, a highly demanding field. Through deep optimization for vulnerability discovery and patch remediation, the model has learned to identify and fix complex code vulnerabilities across 20 programming languages while internalizing a defender-first security mindset.

Developers: Access 3.8 Flash through Google AI Studio, the Gemini API, or Android Studio. You can also explore agent-first workflows in Google Antigravity or generate UI in Stitch.
Enterprise users: Integrate 3.8 Flash directly into the Gemini Enterprise platform to build enterprise-grade intelligent applications.
General consumers: After subscribing to Google AI Pro or Ultra, use 3.8 Flash in the Gemini App, AI Mode in Google Search, and Google Sheets.
Security defenders: Apply through the Fairwind Program. Once certified, gain priority access to 3.8 Flash Cyber for vulnerability discovery and automated patch remediation.

Exceptional value: Delivers reasoning and coding capabilities approaching or surpassing larger frontier models such as Claude Opus 5 and GPT-5.6 at the low price of the Flash series.
Long-horizon software engineering: Autonomously solves complex engineering problems end to end in long-horizon software engineering benchmarks such as DeepSWE v1.1, ranking among the industry's leaders.
Enterprise-grade professional reasoning: Demonstrates the reliability needed to support autonomous tasks in critical areas including financial analysis, legal workflows, and cross-disciplinary expert reasoning.
Dynamic compute adjustment: Supports flexible reasoning-effort adjustment based on task complexity, investing more compute resources in difficult tasks to maximize performance while reducing Token usage when efficiency is the priority.
Cybersecurity specialization: The Flash Cyber version reaches frontier-level performance in vulnerability discovery and automated patch remediation at a much lower cost than comparable large-parameter models.
Native multimodality and agent capabilities: Built on long-running agentic loops, it supports text, code, image, video, and other modalities, and can autonomously iterate to complete complex multi-step tasks.

Official project website: https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/

Autonomous software development: Completes complex code refactoring, bug fixes, and feature development end to end, supporting long-horizon software engineering tasks.
Financial quantitative analysis: Automatically parses financial reports and market data to generate investment research reports, risk assessments, and compliance recommendations.
Legal document review: Quickly analyzes contract clauses, case law, and regulations to assist lawyers with due diligence and compliance reviews.
Cybersecurity defense: Automatically scans multilingual codebases for vulnerabilities and generates remediation patches, helping security teams respond to threats quickly.
Multimodal intelligent applications: Uses natural language to generate 3D interactive games, data visualization applications, and rich-media content.