AI Pulse · Frontier Labs Watch

Google has launched Gemini 3.8 Flash with unusually clear pricing through the end of 2026. It costs $0.75 per million input tokens and $3.75 per million output tokens, then rises to $1.50 and $7.50 respectively from January 2027. For operators building AI into customer support, internal knowledge tools, reporting, or software workflows, that gives you something rare in this market, a forward view of likely model costs rather than a price that could change next month.

The more interesting detail is the Cyber version. Google is restricting it to its Fairwind Program, aimed at government, critical infrastructure organisations, and software maintainers. That tells you this is not being positioned as a general chatbot feature. It is intended for work where finding and fixing security issues has real operational consequences. Chrome Security reported 2.6 times more correct vulnerability patches than larger rival models. Wiz reported 7.5 to 9.7% higher recall on its internal penetration-testing benchmark, at 2.3 to 5.2 times lower cost. Those are independent third-party results, not just Google’s own benchmark claims.

For most businesses, the immediate lesson is not to wait for access to this specific model. It is to treat AI-assisted security and code review as a serious operating area, especially if you maintain software, handle sensitive customer data, or rely on critical systems. This is the kind of thing we build into an AI command centre, where new model capability is tracked against cost, risk, access, and the workflows it could improve.