Google launched Gemini 3.5 Flash at its I/O 2026 developer conference today, and the scale of the rollout is notable. This is not a product for power users or developers only. It is now the default model inside Google Search's AI Mode and the Gemini app globally, reaching over 900 million monthly users.
The practical meaning: when anyone opens Google Search and uses AI Mode, or opens the Gemini assistant on their phone, they are now running on this new model. Businesses whose staff rely on Google Workspace, Gmail summaries, or Gemini for document work are already in this upgrade whether they chose it or not.
The performance case Google makes is actually interesting. A "Flash" model in Google's naming is supposed to be the faster, lighter version, not the top of the range. Yet Gemini 3.5 Flash outperforms Gemini 3.1 Pro, which was their flagship as recently as February, across coding and multi-step task benchmarks. Google says it runs four times faster than comparable frontier models from rivals. The logic is that faster, cheaper models deployed at huge scale are more useful than slower, expensive ones reserved for specialists.
The shift in what AI is being asked to do is also clear here. Gemini 3.5 Flash is designed for what the industry calls "agentic" work: AI that plans, executes, and completes tasks over multiple steps rather than just answering a single question. Google's new personal agent, Gemini Spark, runs on this model and operates around the clock. It can send emails, add calendar events, and handle workflows across your apps while you are not watching. Shopify is using it for merchant data analysis, Xero for tax form preparation, and Macquarie Bank is piloting it for customer onboarding.
Now, the pricing. This is where it gets uncomfortable for anyone paying for AI access. The new 3.5 Flash costs $1.50 per million input tokens and $9 per million output tokens. That is three times the price of the previous Gemini 3 Flash Preview. Compared to the even cheaper Flash-Lite model before that, the price is six times higher. Google is running it for free in its consumer products, but the bill lands on businesses using the API or enterprise plans.
This is not just a Google story. OpenAI's GPT-5.5 launched at double the price of the version before it. Anthropic's Claude Opus 4.7 costs roughly 1.5 times more than its predecessor. All three major AI providers are raising prices on their newest models at the same time, even as older and cheaper models remain available. The bet they are all making is that improved capability justifies the increase, and that businesses will pay it.
For practical decision-making: the older, cheaper Gemini Flash models are not going away immediately, so there is no forced migration today. But the direction is clear. Each new generation arrives at a higher price point, and the new models get integrated into the products people actually use day to day. If you are a business running AI-assisted workflows, audit what model you are currently on and what it costs. And if your team uses Google's free tools, they are already working with this new model, whether or not your IT department has made any decisions about it.