Google's New AI Works Harder, but Your Bill Might Grow

Google just released Gemini 3.8 Flash, a new AI model that the company says "works harder" than its older version. In plain terms, it spends more time reasoning through hard questions and uses tools like search or code-execution in multiple steps instead of giving a quick answer. For developers and companies that rent AI by usage, that sounds helpful — but it may also mean higher costs.

Here's the catch: the per-token price hasn't changed. But since the model spits out more tokens per task (about 30% more), researchers say the total cost per job has jumped roughly 40%. Independent testers still call it the cheapest model at its intelligence level, so you're getting more brainpower for the money — just not as cheap as before. If you subscribe to Google AI Pro or Ultra, you'll likely notice your usage goes up too.

The extra effort pays off in useful places. Gemini 3.8 Flash beats its rivals on tests for software engineering, financial analysis, and legal work — the kinds of jobs where thinking twice actually matters. It can handle "agents," or AI that takes actions for you, more reliably. There's even a special version called 3.8 Flash Cyber, limited to governments and approved partners, meant to find and fix security holes in critical systems. Google says that version comes with safeguards to avoid misuse.

The model is available now for Google AI Pro and Ultra subscribers, plus developers and businesses. If you're an everyday user, you won't notice the difference day to day. But if you're paying for AI by usage—or building tools on top of it—the "harder-working" model might start showing up on your invoice.