Google Cuts Free Gemini to Flash-Lite and Moves Pro Behind the $19.99 AI Pro Plan
TL;DR
Google updated its Gemini Apps limits page to say that, starting October 9, personal accounts without an AI subscription will get only the Flash-Lite model in Gemini. Free users lose Flash and Pro. Subscribers on the cheaper AI Plus tier keep Flash-Lite and Flash but lose Pro, on a date Google will send them by email. The Pro model and the Deep Think reasoning mode now sit behind Google AI Pro at $19.99 a month and above. If you have been treating the Gemini app as a free Pro-class model, you have five days.
What changes on October 9
Google's own support document says: "These changes will start to take effect for users without an AI subscription on October 9th." In its model table, the "Without a plan" column checks only Flash-Lite. Flash and Pro are unchecked.
According to 9to5Google, which spotted the change, the models free users lose are 3.6 Flash and 3.1 Pro, and the one they keep is 3.5 Flash-Lite. Google's document describes Flash-Lite as an efficient model for fast, everyday work and points Pro at more complex math and coding, per tbreak's reading of the same page.
For AI Plus, the language is softer on timing but identical on substance: "For users with an AI Plus subscription, you should receive an email that explains when these changes will take effect for you." After that date, Plus gets Flash-Lite and Flash. No Pro.
The sweetener: Deep Think drops a tier
The same document adds: "If you have an AI Pro or AI Ultra subscription you'll also have access to the Deep Think option for maximum parallel reasoning." Before this change, Deep Think was an Ultra-only feature, which starts at $99.99 a month on Google's plans page. So the $19.99 tier gets better on the same day the tiers below it get worse.
That is a textbook pricing move. You do not raise the price of the plan you want people to buy. You make everything under it noticeably thinner and give the target plan one shiny new thing. Think of a coffee shop that quietly shrinks the small cup to an espresso shot and starts putting a free pastry with the medium.
9to5Google also reports that Gemini will add low, medium, and high effort settings for each available model. Higher effort gives more thorough answers and burns more of your allowance, which matters because of how Google now meters usage.
Limits are compute, not message counts
Model access is only half the picture. Google's support page describes usage limits that are compute-based and refresh every five hours, and it frames each paid tier as a multiple of the free allowance:
- No plan: "standard limits"
- AI Plus: 2x standard
- AI Pro: 4x standard
- AI Ultra: 5x or 20x higher than AI Pro, depending on the subscription
Per tbreak, the five-hour windows roll up into weekly limits, and prompt complexity, features used, and chat length all affect consumption. The shift to compute-based limits started in May, according to both tbreak and 9to5Google. Google publishes no absolute number for "standard," so every multiplier is relative to a baseline you cannot see. Very on-brand for a quota system.
Why Google is doing this now
Google has not published a reason beyond the support page itself. The reporting around it points to two things. Android Headlines reads it as Google "nudging regular users toward its higher-tier plans" by moving Pro exclusively behind the Pro and Ultra paywalls, and calls losing Pro on Plus a sting because it "was easily the best bang-for-your-buck feature on that tier."
The other read is compute. Serving a Pro-class reasoning model to an unpaid audience at Google's scale is expensive, and the company has a new frontier model, Gemini 4 Argon, that it is currently rationing to cyber defenders rather than shipping to the app. Every free Pro query is capacity not spent elsewhere. Our inference, not Google's statement: when a lab with its own TPUs starts metering its free tier this hard, the unit economics of free reasoning have not gotten as cheap as the per-token price charts suggest.
What this means if you build or sell on AI
Free consumer chat is becoming a demo, not a product
For two years the default consumer pitch was "the good model, free, with limits." Google is moving to "a small model, free, and the good one is $20." If your own app rides on users having a free frontier assistant on the side, assume that assumption gets weaker.
Price anchoring is the playbook
If you run a SaaS with tiers, this is worth studying. Google did not touch the $19.99 price. It removed the flagship model from the two tiers below, added a premium feature to the target tier, and introduced effort settings that make heavy users burn quota faster. That combination pushes upgrades without a headline price increase.
Watch the AI Plus emails
AI Plus subscribers do not have a fixed date yet. If you or your team pay for Plus mainly for Pro access, the math changes the day that email lands. The developer API is a separate product with its own pricing and is not part of this support-page change.
Key Takeaways
- From October 9, Gemini users without a subscription get only Flash-Lite. Flash and Pro are removed, per Google's support page.
- AI Plus keeps Flash-Lite and Flash but loses Pro on a date Google will communicate by email.
- Pro and the Deep Think mode now start at Google AI Pro ($19.99/month); Deep Think was previously Ultra-only.
- Limits are compute-based, refresh every five hours, and scale 2x (Plus), 4x (Pro), and 5x or 20x Pro (Ultra) over an unquantified baseline.
- Low, medium, and high effort settings are coming, and higher effort consumes more allowance.
- For builders: treat a free frontier assistant as a shrinking assumption, and note the tier-thinning pricing play.
Sources: Google Gemini Apps Help: model access and limits, 9to5Google, Android Headlines, tbreak, Google AI plans