Google has two help pages about Gemini's free-tier model access, and they don't agree

What was reported, starting 3 October

9to5Google reported on 3 October 2026 that Google had published a help page announcing a restructuring of Gemini app model access, effective 9 October. From there it spread fast: TechRadar, Computing.co.uk, and dozens of smaller blogs repeated the same shape of claim: free accounts limited to Flash-Lite only, AI Plus losing Pro on a rolling per-account schedule, and AI Pro gaining Deep Think, a reasoning mode previously exclusive to AI Ultra.

Two Google pages, two different answers

Gemini Apps Help turns out to have two overlapping pages on this exact topic, and they disagree. Answer 16275805, "Gemini Apps limits & upgrades for Google AI subscribers," is a general reference page with no visible date, no mention of an October change, and a Model Access table that currently shows every tier, including "Without an AI Plan," with Yes across Flash-Lite, Flash, and Pro. Answer 17004136, "Changes to Gemini model access and limits," is explicitly dated: "Starting in October 2026, there will be changes to model availability for Gemini Apps when you use a personal account," taking effect for free accounts on 9 October and for AI Plus accounts on a schedule sent by email.

We checked both pages directly. Google's general reference page (16275805) is the easier one to find and reads like current truth; it isn't. The page actually written for this change (17004136) says something different for the free and AI Plus tiers. Google hasn't explained why the two pages disagree; the simplest explanation is that 16275805 is older and hasn't been updated to reflect the October change described on 17004136, but that's our inference, not something Google states.

The part every report flattened: "Auto," not a hard lockout

The dated page's own wording matters here. For free accounts: "Auto will be the default setting for users without a Google AI plan. Gemini will use auto model selection to send each prompt to the Gemini model that suits it best. Most prompts will go to Flash-Lite, our fastest model. Prompts that need deeper reasoning may go to Flash or Pro." Only the next line describes an actual lockout: "If smart model selection is turned off, chats for users without an AI plan will use Flash-Lite for all responses."

That's a meaningfully different claim than "free users are restricted to Flash-Lite," the framing nearly every outlet used. As published, a free account that leaves Auto on can still have a prompt routed to Flash or Pro when Gemini judges the prompt needs it; the restriction is a default and a ceiling on manual control, not a wall around Flash-Lite specifically. Turning Auto off is the thing that actually locks you to Flash-Lite, and that's a setting change you'd make yourself, not something Google silently does to every free account.

What AI Plus actually loses

The dated page's table is unambiguous for AI Plus: a checkmark under Flash-Lite, a checkmark under Flash, and an X under Pro. AI Pro and AI Ultra both show checkmarks across all three models. That part of the original reporting holds up exactly as described, AI Plus subscribers lose Pro access, on a timeline Google says will arrive by email rather than on a single fixed date.

What we couldn't confirm is the Deep Think claim. Several October reports said AI Pro would gain Deep Think, a reasoning mode previously exclusive to AI Ultra. Neither Google page we checked states that directly: the dated change page (17004136) doesn't mention Deep Think at all, and the undated general page (16275805) still lists it as "Deep Think (AI Ultra only)." It's possible that part of the story is accurate and just undocumented on these two pages, or that it was an early plan that didn't ship; we don't have a Google source confirming it either way, so treat AI Pro gaining Deep Think as unconfirmed rather than denied.

What this means if you use Gemini Notebook

If you came here looking for Gemini Notebook (formerly NotebookLM) specifically: this story concerns the general Gemini app chatbot at gemini.google.com, not Gemini Notebook. The two run on different model tracks. Gemini Notebook's documented chat model is Gemini 3.5, separate from the Gemini 3 Flash-Lite/Flash/Pro lineup this help page covers, and Google has never tied Gemini Notebook's source limits or Audio Overview quotas to the Gemini app's per-message model routing. Your AI Plus, Pro, or Ultra subscription still gates Gemini Notebook's own source counts and feature set the way it always has; this specific change doesn't touch that. The subscription tier names overlap, Plus/Pro/Ultra gate both products, but a change to one doesn't automatically apply to the other, which is worth remembering any time you read about a Google AI plan change and wonder whether it reaches Gemini Notebook too.

How to check your own access

• Read the dated page directly: support.google.com/gemini/answer/17004136 is the one Google actually wrote for this specific change. The other Gemini Apps help page covering limits is a general reference that, as of this writing, hasn't been updated to match it.
• Check your model picker. If you're on a free account, look at whether Auto is on or off in your settings. With Auto on, you're not locked to Flash-Lite, Gemini still routes complex prompts to Flash or Pro; turning Auto off is what restricts you.
• If you're on AI Plus and still see Pro in your picker, you likely haven't reached your account's rollout date yet; Google says the AI Plus timing arrives by email rather than on a fixed calendar date.
• Don't assume either Google page is permanent. Both state limits and model availability "may change," and Google has revised compute-based limits before (see the May 2026 adjustments below). Re-check the dated page yourself before planning a workflow around a specific access level.

Google has adjusted Gemini limits under pressure before

This wouldn't be the first time Google tightened Gemini's compute-based limits and then adjusted after pushback. In late May 2026, after the original 17 May compute-based limit system left users hitting caps faster than expected, Google Labs VP Josh Woodward posted a set of fixes on X: capping how much quota a single large Gemini 3.1 Pro request could consume, making Flash-Lite prompts free and uncounted against quota, and no longer charging quota for failed requests. Whether this October change sees a similar adjustment isn't something we can predict; it's precedent for Google revising a restrictive rollout within weeks, not proof that this one will follow the same path.

People also ask

Did Google actually restrict free Gemini users to Flash-Lite on October 9, 2026?

Not exactly as reported. Google's dated help page (support.google.com/gemini/answer/17004136) confirms free accounts moved to Auto model selection on 9 October 2026, which routes most prompts to Flash-Lite but can still send complex prompts to Flash or Pro. Only manually turning Auto off locks an account to Flash-Lite for every response.

Did AI Plus lose access to Gemini Pro?

Yes, confirmed on Google's dated change page: AI Plus shows checkmarks for Flash-Lite and Flash but an X for Pro. The timing varies by account; Google says AI Plus subscribers receive an email with their specific effective date.

Did AI Pro gain access to Deep Think?

Unconfirmed. Neither Google help page we checked states this directly. The dated change page doesn't mention Deep Think at all, and Google's other, undated Gemini Apps reference page still lists Deep Think as Ultra-only.

Why do Google's own pages disagree about this?

Google maintains two overlapping Gemini Apps help pages: a dated one specifically written for the October 2026 change (answer/17004136), and a general, undated reference page (answer/16275805) that, as of this writing, still shows the pre-change access levels. Google hasn't explained the discrepancy; the likely explanation is that the general page simply hasn't been updated yet.

Does this affect Gemini Notebook (NotebookLM)?

No. This story concerns the general Gemini app chatbot at gemini.google.com. Gemini Notebook runs on a separate model track (Gemini 3.5, not the Gemini 3 Flash-Lite/Flash/Pro lineup covered here) and its own source and feature limits, which this change does not touch.