# Google releases Gemini 3.8 Flash and a cybersecurity model — QANATONE > Google updated its Flash line: 3.8 Flash improves at coding and agent tasks, and 3.8 Flash Cyber is limited to trusted defenders. Kaynak: https://www.qanatone.com/en/haber/gemini-3-8-flash/ --- [News](https://www.qanatone.com/en/haber/) # Google releases Gemini 3.8 Flash and a cybersecurity model 2 Sep 2026 1 min read News · Models Google updated its Flash line: 3.8 Flash improves at coding and agent tasks, and 3.8 Flash Cyber is limited to trusted defenders. ## What changed? Google announced Gemini 3.8 Flash on September 2, describing it as its most intelligent workhorse model. According to Google, it improves significantly on 3.7 Flash in software engineering, agentic tasks and multi step reasoning in specialized domains. It is the latest in a run of quick Flash updates; 3.7 Flash became generally available on August 13. ## Where and at what price? 3.8 Flash went live in the Gemini app for Google AI Pro and Ultra subscribers, in AI Mode in Google Search and in Gemini in Google Sheets. API pricing stayed at the same introductory level as 3.7 Flash: $0.75 per million input [tokens](https://www.qanatone.com/en/nedir/token-nedir/) and $3.75 per million output tokens. ## The cybersecurity version On the same day Google introduced Gemini 3.8 Flash Cyber, presented as its most capable cybersecurity model, strong at finding vulnerabilities and generating patches automatically. Because of misuse risk it is available only to trusted defenders admitted to the new Fairwind Program. The approach mirrors OpenAI keeping [GPT-6 Astra](https://www.qanatone.com/en/haber/gpt-6-astra-agi-cagi/)'s high risk cyber abilities under tight control at launch. ## Why it matters The Flash series is Google's line of fast, cheap models. Much everyday work, such as classification, summarizing and first replies to customer messages, runs on models like these. When capability rises while price stays the same, harder jobs can move into this tier on the same budget. Our [inference](https://www.qanatone.com/en/nedir/inference-nedir/) explainer covers why that matters for cost. ## More articles [News · Models ### Claude Fable 5.1 launches; Mythos 5.1 stays restricted 1 Sep 2026 · 1 min](https://www.qanatone.com/en/haber/claude-fable-5-1-mythos-5-1/)[News · Ads ### ChatGPT ads hit a $1 billion annual run rate in under 200 days 31 Aug 2026 · 1 min](https://www.qanatone.com/en/haber/chatgpt-reklamlari-1-milyar-dolar/)[News · Funding ### a16z raises a $1.1 billion fund for the physical side of AI 28 Aug 2026 · 1 min](https://www.qanatone.com/en/haber/a16z-machine-age-fonu/)[News · Payments ### Coinbase x402: AI agents made 14 million payments in 30 days 27 Aug 2026 · 1 min](https://www.qanatone.com/en/haber/coinbase-x402-ajan-odemeleri/)[News · Models ### Ask Gemini arrives in Google Chat as a single command line 26 Aug 2026 · 1 min](https://www.qanatone.com/en/haber/google-chat-ask-gemini/)[News · Funding ### AI agent Instinct, built by a 23-year-old, hits $2.5 billion 26 Aug 2026 · 1 min](https://www.qanatone.com/en/haber/instinct-2-5-milyar-dolar/)[News · Infrastructure ### Nvidia reports $96.2 billion in quarterly revenue 26 Aug 2026 · 1 min](https://www.qanatone.com/en/haber/nvidia-96-milyar-dolar-ceyrek/)[News · Funding ### Salesforce's Agentforce passes $1.5 billion in ARR 26 Aug 2026 · 1 min](https://www.qanatone.com/en/haber/salesforce-agentforce-1-5-milyar/)