Google released Gemini 3.8 Flash on Wednesday, September 2, 2026, its fourth Flash model in four months and an attempt to answer the past summer's steady drip of Chinese open-weights releases that had knocked Google out of the top of the intelligence charts. The model matches or beats several higher-cost frontier systems on public benchmarks while holding an introductory price of $0.75 per million input tokens and $3.75 per million output tokens.
Cheaper agents, longer horizons
Set to high reasoning, Gemini 3.8 Flash scores 59 on the Artificial Analysis Intelligence Index, a three-point jump on Gemini 3.7 Flash and level with GPT-5.6 Sol and Grok 4.6. At $0.58 per Intelligence Index task, Artificial Analysis calls it the cheapest model at its intelligence level, roughly six times less expensive per task than Claude Fable 5.1. Google's own numbers show the model spending about 40 percent more tokens per task than 3.7 Flash because it runs additional reasoning steps and tool calls at higher effort settings.
Fairwind Cyber for governments and maintainers
Alongside the general release, Google launched the Fairwind Program to give governments, critical-infrastructure organisations and software maintainers access to Gemini 3.8 Flash Cyber, a variant tuned for software vulnerability discovery and remediation. The move mirrors the cyber-restricted access model Anthropic uses for Claude Mythos 5.1, and lands the same week OpenAI rated its next model as a critical cyber risk under its Preparedness framework.
Where it runs
Gemini 3.8 Flash is available through Google Antigravity, the Gemini API in Google AI Studio and Android Studio, and Google's Stitch design service. Enterprises reach it via Gemini Enterprise; consumers on AI Pro and Ultra subscriptions get it inside the Gemini app, AI Mode in Google Search and Gemini in Google Sheets. The Cyber variant sits behind Fairwind's application review process.
Reporting based on coverage from The Register, Neowin, Google, and Artificial Analysis.