Google will cut free users of the Gemini app down to the Flash-Lite model from October 9, removing access to Flash and Pro, according to an updated support document reported by 9to5Google on October 3. Paid tiers change too: AI Plus subscribers at $4.99 a month keep Flash-Lite and Flash but lose Pro, while Deep Think, the maximum parallel reasoning option so far reserved for the $99.99 AI Ultra plan, moves to the $19.99 AI Pro tier.
The shift lands on top of a compute-based usage limits change in May, and it is the second time this year Google has redrawn the line between what casual users get and what subscribers pay for. The support page frames it as a reallocation, not a downgrade, and points out that paid tiers above Plus see no changes at all.
| Plan | Flash-Lite | Flash | Pro | Deep Think |
|---|---|---|---|---|
| Without a plan | Yes | Removed October 9 | Removed October 9 | No |
| AI Plus, $4.99 | Yes | Yes | Removed | No |
| AI Pro, $19.99 | Yes | Yes | Yes | New |
| AI Ultra, $99.99 | Yes | Yes | Yes | Yes |
Who is affected
The free tier is the visible cut. Anyone using Gemini without a subscription keeps only Flash-Lite, the lightest model in the lineup, from October 9. Flash and Pro both disappear from the free app. For a user who has been leaning on the free version for study help, drafting or coding, the practical change is a slower, shallower model on the same interface.
AI Plus, the $4.99 tier, keeps the middle of the lineup but loses Pro outright. These subscribers will receive an email explaining when the change lands on their account, per the support document. That is the group losing the most per dollar paid, since Pro was until now part of the Plus bundle.
AI Pro subscribers at $19.99 get a release worth having. Their plan gains Deep Think, Google’s maximum parallel reasoning mode, which until now has been locked to the $99.99 AI Ultra tier, alongside top-tier Flash and Pro access. In effect, Google is using Deep Think as the new flagship benefit to justify the Pro tier, now that Pro-model access alone has migrated down the pricing ladder.
Why Google is doing this now
Two numbers explain the timing. Inference compute for a frontier model costs real money per query, and a free tier serving hundreds of millions of sessions on Pro-class models is a bill with no offsetting revenue. The May change began metering usage. The October change cuts the default model on the non-paying tier entirely. Both narrow the gap between what users consume and what they fund, and both follow the same playbook Claude and ChatGPT ran earlier: free tiers get the economy models, reasoning gets priced.
The second driver is the Gemini 4 Argon launch. Google announced its new frontier model earlier this week, with first access going to AI Ultra subscribers, and has yet to say whether Argon will be classified as a Pro model or sit in a tier of its own. Freezing the experience ladder now gives Google a clean slate to slot the new flagship wherever it wants without surprising a different cohort of subscribers mid-cycle.
There is a competitive read too. AI Weekly’s October 4 coverage of the OpenAI beat shows the same pattern across labs: flagship access separating from the free tier, reasoning features defending premium pricing, and usage limits tightening as compute budgets get honest. Gemini’s move is the Google expression of an industry-wide repricing, not a lone experiment, and it arrives while rivals charge $20 or more for their own mid tiers.
Effort levels arrive in October
The same window adds a small but practical feature. The Gemini app looks set to gain low, medium and high effort levels for each available model. Higher settings push the model to spend more compute on a task, give longer and more thorough answers, and also use more of the user’s limit. Until now the app has offered only an extended thinking toggle for complex problem solving. The three-step ladder matches what is already available in Google AI Studio and the Antigravity IDE, so the app is catching up to the rest of Google’s own tooling rather than introducing something new.
For users who track such things, effort levels matter more than headlines suggest. A low-effort run on Flash will handle a summary or a rewrite for a fraction of the quota, and reserving high effort for genuinely hard questions stretches a paid plan much further. The feature rewards users who pay attention to which model and which effort tier they invoke, and it gives Google another dial to tune as costs shift.
What it means for different users
Casual users lose the most in relative terms, and they will notice within days of October 9. Students, hobby coders and anyone drafting on the free tier will sit on Flash-Lite, which handles short prompts fine and longer or technical ones noticeably worse. Some will accept the slower model. Others will try a rival’s free tier, and OpenAI and Anthropic both still give free users a stronger default model than Flash-Lite Today.
AI Plus subscribers face a harder call. The tier’s value proposition was Pro access at a budget price, and that is gone. The email Google will send explains the timing, but not a discount, so Plus holders will effectively be deciding whether Flash is worth $4.99 as a standalone product.
AI Pro subscribers at $19.99 come out ahead on paper. They gain Deep Think, which the $99.99 plan charged a fivefold premium for, with no other change to their plan. Anyone using Gemini for serious work who was already on Pro got a free upgrade in reasoning depth.
Google has not said whether Pro access will ever return to the free tier, and the support document leaves the door open on future reallocations. For now, October 9 is a clear deadline, and Pro-class reasoning on Gemini now starts at $19.99 a month.