Google Debuts Gemini 3.8 Flash and a New Cybersecurity Model

Google has announced Gemini 3.8 Flash, its latest Flash model, with improvements in coding, agentic tasks, reasoning, and several specialist benchmarks.

The launch comes only several weeks after Gemini 3.7 Flash arrived with improved coding capabilities, better agent performance, and lower starting prices.

Gemini 3.8 Flash starts at $0.75 per million input tokens

Google has set promotional pricing for Gemini 3.8 Flash at $0.75 per million input tokens and $3.75 per million output tokens.

Those prices will remain available through the end of 2026. Starting January 1, 2027, the rates will double to $1.50 per million input tokens and $7.50 per million output tokens.

Google also warns that the model can consume more output tokens on difficult tasks. Gemini 3.8 Flash may perform additional reasoning steps and make repeated tool calls when handling complex requests, increasing total token usage.

Gemini 3.8 Flash closes the gap with frontier models

As for benchmarks, Google reports a substantial improvement in software engineering performance.

Gemini 3.8 Flash scored 71% on DeepSWE v1.1, compared with 65.3% for Gemini 3.7 Flash. Claude Opus 5 remained ahead with a score of 74%.

The new Gemini model performed even better on Terminal-bench 2.1, reaching 89.4%. That result put it slightly ahead of Claude Opus 5 and GPT-5.6 Sol.

Google also reported strong results across finance, legal reasoning, chart understanding, video understanding, and biology benchmarks.

Google also launches Gemini 3.8 Flash Cyber

Alongside the general-purpose model, Google introduced Gemini 3.8 Flash Cyber, which focuses on vulnerability discovery and automated software patching.

The cybersecurity model scored 47.2% on CWE-Bench, narrowly behind the 47.8% result achieved by a leading frontier model.

Google says its internal tests produced another notable result. Gemini 3.8 Flash Cyber generated 2.6 times more correct patches for Chrome vulnerabilities than larger commercial models.

Access to the cybersecurity model remains restricted. Google is offering it through its Fairwind Program to trusted government agencies, critical infrastructure operators, and software maintainers.

Gemini 3.8 Flash is available now

Developers can access Gemini 3.8 Flash through the Gemini API, Google AI Studio, Android Studio, Antigravity, and Gemini Enterprise.

Google AI Pro and Ultra subscribers can also use the model through the Gemini app, AI Mode in Google Search, and Gemini in Google Sheets.

The launch comes as competition between major AI developers continues to expand into specialized enterprise and cybersecurity workloads.

In other AI news, OpenAI has launched ChatGPT Mil for secure military environments, while ChatGPT also gained improved Epic EHR access for healthcare workflows.

OpenAI has also detailed its Astra cybersecurity model ahead of its planned release.