Opening the Gemini app on October 9 meant something different for people who had never paid Google a cent. The heavier models were gone. Free users woke to a single option, Gemini 3.5 Flash-Lite, the lightest and cheapest model in the lineup. The change arrived quietly, through an update to a support document, but it redrew the deal Google offers to the hundreds of millions of people who use its assistant for free.
Before October 9, free accounts could dip into Gemini 3.6 Flash and 3.1 Pro with limited usage. Both were removed from the free tier. The $4.99-a-month AI Plus plan, one step up, keeps Flash-Lite and Flash but loses Pro. The $19.99 AI Pro plan keeps all three models and gains something new: Deep Think, a reasoning mode that had been reserved for the $99.99-and-up tiers.
Deep Think is Google’s answer to the reasoning-style models that have become the industry’s most expensive product. The company describes it as “maximum parallel reasoning,” a mode that spends more compute to chew through hard problems. Handing it to $19.99 customers, rather than keeping it for the $99.99 and $199.99 plans, is a price cut dressed as a perk.
Alongside the tier reshuffle, Google added three reasoning intensity levels, low, medium and high, across the models. The setting lets a user trade speed for depth on a per-prompt basis. It is the same dial the industry has converged on, a way to meter the compute that thinking burns, and it gives paying users finer control over how much of their allowance each prompt consumes.
The lineup has settled into a clear three-rung ladder. Flash-Lite is the workhorse, built to run fast and cheap on routine tasks. Flash is the general-purpose mid-tier, tuned for everyday questions and generation. Pro is the flagship, reserved for the hardest reasoning and the longest documents. Until October 9, free users could sample the top of the ladder; now the ladder starts at a paid rung.
Reasoning modes are the industry’s newest cost center. A model that thinks before it answers can burn several times the compute of a straight response, which is why the feature has been rationed. By giving Deep Think to the $19.99 tier and adding intensity dials, Google is inviting paying users to spend their allowance on quality rather than quantity, and quietly conditioning them to think of reasoning as a metered good.
The economics are the story. Free AI is expensive to serve. Flash-Lite is a cost-efficient model Google has positioned for high-volume, low-margin work: translation, simple data processing, and the agentic chores that must run cheaply at scale. Pushing free users onto it cuts the bill for serving a tier that generates no revenue.
Google is not alone in tightening the free tier. The industry spent two years giving away frontier models to win users, and the bill has come due as those models got more expensive to run. Rivals have made the same move: the free tier gets the lightweight model, and the heavy reasoning goes behind a paywall. Google’s version is sharper. A free tier reduced to a single model is a harder line than most competitors have drawn.
The middle tier lands in an awkward spot. At $4.99, AI Plus now loses Pro, the model it once used as a selling point, while gaining only Flash. The distance between $4.99 and $19.99 has widened, and the cheapest plan that keeps all three models is now AI Pro. For a user who wants Pro only occasionally, the jump is steep, and the value of the cheap plan is suddenly harder to explain.
The change also hints at how Google wants to meter its frontier models. The 3.1 Pro model, the top general-purpose option, is now effectively a paid product. Flash, the mid-tier, sits at the $4.99 threshold. Flash-Lite, the workhorse, is the loss leader that keeps free users in the app and feeds them toward a subscription. Each rung now corresponds to a clear model, and each model to a clear price.
For Google, the free tier has always been a funnel rather than a business. The assistant is the front door to Gemini’s paid plans and to the models’ use across Google’s other products, from Search to Workspace. A thinner free tier raises the pressure on that funnel: users who relied on Flash or Pro for free now face a choice between paying and leaving.
The decision also lands in the middle of a price war. Microsoft this week launched a decision-scoring model whose pricing was nearly identical to a rival from the startup TypeSafe AI, and the reasoning market is crowded with aggressive undercutting. Google’s move to gate its best models behind subscriptions runs against that current, betting that quality and brand can hold a line that rivals keep lowering.
Whether the squeeze works is an open question. Free users have alternatives, and the cost of switching an AI assistant is measured in minutes, not months. Google is wagering that Deep Think at $19.99 is enough to pull people up the ladder, and that the free tier’s vast audience will tolerate a single lightweight model. The next few weeks will show which side of that bet is right.
