Google will restrict the models available to Gemini app users without a paid plan from October 9, 2026, leaving them with a single model, Flash-Lite, according to a Gemini Apps Help page that 9to5Google reported on October 3, 2026. Subscribers to the $4.99-a-month AI Plus plan keep two models and lose the third, while the $19.99-a-month AI Pro plan gains a reasoning option, Deep Think, that is currently confined to the two most expensive plans.

In Short

Google is changing which Gemini AI models people can use in the Gemini app, depending on what they pay. From October 9, 2026, people who pay nothing get only the entry-level model, people on the $4.99 AI Plus plan lose the strongest one, and the $19.99 AI Pro plan adds a deeper-thinking mode that used to sit on the most expensive plans only. In practice, the same question typed into the app can now be answered by a different model depending on the plan behind the account.

What Google's help page says

The primary source is a help-centre page titled "Changes to Gemini model access and limits". It opens with a notice that, from October 2026, model availability in the Gemini app will change for people signed in with a personal account. According to the page, the changes begin on October 9 for users without an AI subscription, while AI Plus subscribers are due to receive an email stating when the restrictions reach them. The captured copy carries no revision date; 9to5Google, in a report timestamped 1:57 pm Pacific on October 3, said the document had been updated that week.

Combining the page's two tables gives the following picture. Prices are absent from Google's page and come from 9to5Google.

PlanMonthly priceFlash-LiteFlashProDeep ThinkUsage limit
No planFreeYesNoNoNoStandard limits
AI Plus$4.99YesYesNoNo2x higher than standard
AI Pro$19.99YesYesYesYes4x higher than standard
AI Ultra$99.99 or $199.99YesYesYesYes5x or 20x higher than AI Pro

The page identifies the models by family name only - Flash-Lite, Flash and Pro - and gives no version numbers. 9to5Google adds them in parentheses: 3.5 for Flash-Lite, 3.6 for Flash and 3.1 for Pro. If those labels hold, accounts without a subscription lose Gemini 3.6 Flash, which Google released on July 21, 2026, with the Gemini app listed among the surfaces open to all users. Eighty days will separate that release from the restriction.

AI Plus subscribers face a narrower change. These customers pay $4.99 a month for a plan that, as of June, also bundles 400 GB of storage and 200 Google Flow credits. Flash-Lite and Flash remain; Pro goes. According to 9to5Google, they will "receive an email that explains when these changes will take effect", so the date for that plan is communicated individually rather than printed on the page. AI Pro and AI Ultra subscribers keep all three models in both documents. The one addition on those plans is Deep Think, covered below.

Housekeeping details fill out the notice. Google asks users to install the latest version of the Gemini mobile app, with update instructions for Google Play and the App Store, and states that a plan can be started, upgraded, changed or cancelled at any time. The notice is confined to personal accounts; work and school accounts are not addressed.

Effort settings, Deep Think and the usage meter

Beyond model access, the page describes a second dial. For each available model, an effort level of low, medium or high can be selected; higher settings raise the model's ability to complete tasks and produce more thorough answers, at the price of consuming more of the user's limit. The two documents differ on timing. Google's tip is written in the present tense, whereas 9to5Google says the app "looks to be adding" the levels in October and, for now, offers only an "Extended thinking: Complex problem solving" option. The three levels would match what Google AI Studio and Antigravity already offer, according to the outlet.

Deep Think is the other control. Google's page describes it as the option for "maximum parallel reasoning" and assigns it to AI Pro and AI Ultra subscribers. According to 9to5Google, it is at present confined to the $99.99 and $199.99 monthly plans, so the $19.99 tier is the one that gains it. Neither document says on which date Pro subscribers receive it.

Both controls sit on top of a meter introduced earlier in the year. The page's section on previous changes records that usage limits and model availability for Gemini app users over 18 changed starting on May 17, 2026, with users under 18 following on July 24, 2026. The limits are compute-based and refresh every five hours until a weekly cap is reached. The calculation weighs the complexity of a prompt, the features in use and the length of a chat, so two requests of similar size can draw down the allowance at different rates. Google names the heaviest consumers of that allowance: media generation covering images, videos and music, Deep Research, the Pro model, and extended thinking, Deep Think included.

Paid plans scale the allowance. AI Plus is listed at 2x higher than standard limits, AI Pro at 4x higher, and AI Ultra at 5x or 20x higher than AI Pro depending on the subscription. Google does not publish the standard allowance itself, so every figure is relative. Read as straight multiples, the Ultra figures compound to 20x or 80x the standard level, though the phrase "higher than" leaves room for another reading, and neither document pairs the two Ultra multipliers with the two Ultra prices.

The commercial backdrop

Neither document offers a rationale for the changes. The context that follows is therefore drawn from earlier coverage, not from anything Google has said about this revision.

Scale comes first. The Gemini app reached 900 million monthly active users in May 2026, and Alphabet put the figure at 950 million for the second quarter. In the United States, Comscore's panel data, published on September 22, 2026, gave Gemini 30% of AI prompt volume in June, up from 17% in January; ChatGPT fell from 70% to 50% over the same months. Neither attached document breaks the Gemini audience down by plan.

That audience does not see advertising inside the app. Dan Taylor, Google's vice president of global ads, denied on December 8, 2025 that ads were coming to the Gemini app and repeated in January 2026 that no plans existed for them, drawing a line between the assistant and the AI answers in Search, where paid formats run. With no advertising in the app, plan structure - $4.99, $19.99, $99.99 and $199.99 a month, according to 9to5Google - is the main commercial lever visible in the two documents. The ladder was itself reshuffled in May: commentary on the I/O keynote summarised by PPC Land recorded the AI Ultra plan starting at $100 a month, down from $250, and a tightening of free usage. OpenAI has taken the opposite route for its rival product; Similarweb's 2026 Generative AI Landscape report found advertising in 26% of US desktop ChatGPT chats during June 2026.

Cost is the other thread. Jon Morra, Zefr's chief AI officer, argued in a column dated August 24 that the era of assuming AI tokens get cheaper every quarter had closed, citing a frontier model that launched at twice the cost of its predecessor and subscription tiers that no longer bundle unlimited usage. Agencies have built ceilings of their own: PMG pools access to major models under a daily per-user token cap, with an escalation review when someone reaches it. On the developer API, Gemini 3.5 Flash-Lite arrived in July priced at $0.30 and $2.50 per million tokens and running at 350 output tokens per second, in a release that also lowered Flash prices as the 3.6 version used 17% fewer output tokens.

Plan-based gating is not new to the app. Gemini Spark, the background agent introduced at Google I/O on May 19, 2026, reached macOS in beta on June 30 for Ultra subscribers in the United States only, and its support pages require a Pro or Ultra subscription. The June 12 rollout of Search agents to every supported language was likewise restricted to Ultra, a sequence PPC Land has described as Ultra first, Pro next and the wider base later.

What remains unclear

Where, then, does the newest model fit? 9to5Google says it is not clear where Gemini 4 Argon will sit. Google announced earlier that week, according to the outlet, that Argon would come first to AI Ultra subscribers, and it has not said whether Argon will be classified as a Pro model or placed in a new tier above it. The help page, which lists only Flash-Lite, Flash and Pro, does not mention Argon. Google had already confirmed on July 21 that pre-training of Gemini 4 had begun.

Version numbers also differ across Google surfaces. AI Mode, Google's conversational search surface, moved to Gemini 3.8 Flash on September 2, after Gemini 3.7 Flash reached AI Pro and Ultra subscribers in English in August. Sources disagree on the default most users receive there; practitioners testing the model picker in August 2026 described it as Gemini 3.5 Flash-Lite, the model that remains for accounts without a subscription. The 3.7 and 3.8 numbers run ahead of the 3.6 that 9to5Google attaches to the app's Flash tier, and Google's page does not say whether the two surfaces track the same version.

Smaller gaps remain in the documents:

  • Deep Think for AI Pro carries no start date.
  • The 5x and 20x Ultra multipliers are not matched to the $99.99 and $199.99 prices.
  • The effort selector is described as available by Google and as forthcoming by 9to5Google.
  • Regions, and accounts other than personal ones, go unmentioned.
  • Subscriber counts by plan are not published.

Why this matters to the marketing community

For anyone measuring how brands appear in AI answers, model identity is a variable that keeps moving. The IAB's AI visibility framework, as PPC Land covered it, gives model drift the longest treatment and illustrates it with a brand whose share of voice shifts five points overnight because a model was retrained. A plan-based gate adds a variation that involves no retraining at all: the same prompt, typed into the same app, can reach Flash-Lite on one account and Pro on another. Which model, then, does a sample of Gemini answers describe? A sample collected through a single paid account captures one slice of what the app produces.

The funding model matters too. Gemini's audience is large and, on Comscore's figures, gaining share, yet the app itself carries no ads, so the experience a person receives is shaped by what that person pays rather than by what an advertiser buys. Cost discipline, meanwhile, is arriving on both sides of the market: PMG caps token use per employee, and Google meters consumer use by compute. Whether the lightest tier becomes the default experience at scale is something neither document answers.

The next data points are dated. October 9 applies to accounts without a subscription, an emailed date applies to AI Plus, an unspecified October window applies to the effort selector, and Argon's classification is still open.

Timeline

Summary

Who: Google, which operates the Gemini app, and the app's users on personal accounts: those with no subscription, AI Plus ($4.99 a month), AI Pro ($19.99) and AI Ultra ($99.99 or $199.99), according to 9to5Google.

What: A revised Gemini Apps Help page restricts accounts without a subscription to Flash-Lite, removes Pro from AI Plus, extends Deep Think to AI Pro, and describes low, medium and high effort levels for each model, on top of compute-based usage limits that refresh every five hours up to a weekly cap.

When: Google updated the page during the week to October 3, 2026, and 9to5Google reported it that day. The restrictions start on October 9, 2026 for accounts without a subscription; AI Plus subscribers are due an email with their own date.

Where: The Gemini app, including its Android and iOS versions, for personal accounts. Neither document names countries or covers work and school accounts.

Why: Neither document gives a reason. Context from earlier coverage points to the compute-based limits introduced from May 17, 2026, industry-wide pressure on AI token costs, and an ad-free app whose plan tiers are the commercial lever visible in the documents.