
Gemini Free Tier: What You Lose on 9 October, and What to Do
From 9 October 2026 the Gemini free tier drops from three models to one. Free accounts keep Flash-Lite and lose Flash and Pro. Google AI Plus subscribers at $4.99 a month lose Pro as well, and AI Pro subscribers gain Deep Think, which until now was reserved for the $99.99 Ultra plan.
None of the models changed. What changed is which of them you are allowed to pick. That makes this a budget decision rather than a prompting one, and it lands three days from now.
Key Takeaways
- From 9 October 2026, a free Gemini account can use only Flash-Lite. Flash and Pro move behind a paid plan.
- Google AI Plus at $4.99 a month keeps Flash-Lite and Flash but loses Pro. AI Pro at $19.99 a month keeps all three and gains Deep Think.
- Google told AI Plus subscribers they “should receive an email that explains when these changes will take effect for you”, so 9 October is the start of a rollout rather than one switchover for everybody.
- Every remaining model gains low, medium and high effort settings, and Google’s own note is that higher effort “will also use more of your limit”.
- Gemini 4 Argon, announced on 30 September with a 1 million token output limit, is not part of any of this. It went to cyber defenders first and has no consumer date.
What Changes on 9 October, Tier by Tier
The model picker shrinks, and it shrinks by plan. According to 9to5Google, which reported the in-app notice on 3 October, free users keep Flash-Lite 3.5 and lose Flash 3.6 and Pro 3.1. AI Plus keeps Flash-Lite and Flash. AI Pro and AI Ultra keep all three.
AI Pro gets something back in the same change: Deep Think, which Google describes as “maximum parallel reasoning” and which can take minutes to answer. That was an Ultra-only feature before 9 October.
Google has published no rationale for the restriction. There is also no indication that this is regional, and no subscriber numbers to tell you how many people it touches. If you want to know whether your own account is affected, the only reliable test is to open the model picker and count what is in it.

Google Shipped Its Best Model to Almost Nobody
Six days before the free tier shrank, Google announced its most capable model and gave it to a handful of security teams. Gemini 4 Argon, announced on 30 September 2026, is described by Google as its “frontier model for real-world coding, enterprise knowledge work, and cyber defense”. It scores 77.9 percent on DeepSWE v1.1, 68 percent on CWE-bench v1 and 91.7 percent on LVBench.
The headline number is a 1 million token output limit, up from 64,000. That is an output limit, not a context window: it is the size of what the model can write in one go, not the size of what you can feed it. It matters when you want a full report in a single response, and not at all when you want to drop a year of campaign data into a chat.
Argon is rolling out “to a set of trusted cyber defenders through our Fairwind Program”, with a version that has no cyber guardrails for those teams and a guarded version for everyone else. Google says broader release starts with paid API customers and AI Ultra subscribers, and gives no date. Introductory API pricing is $2 per million input tokens and $10 per million output, rising to $4 and $20 afterwards. So the practical answer for a content operation is that Argon is not a tool you can plan around this quarter.
What the Gemini Free Tier Can Still Do
Flash-Lite is a working model, not a demo. It handles the volume work most marketers actually run through a chat window: rewriting a caption six ways, pulling the claims out of a press release, turning a transcript into bullet points, drafting a subject line set. The 32,000 token context window on the free plan is not changing, per 9to5Google.
What it will not do is the reasoning work people were quietly using Pro for. Long competitor comparisons, a brief that has to hold ten constraints at once, anything where you were pasting in a spreadsheet and asking for analysis: that is the capability leaving the free plan.
Google is also replacing Gems with skills, which are reusable instructions you build once and call later, and skills do not need a subscription. From 9 October the Gemini free tier will be running them on Flash-Lite. Google has published nothing about how a skill written against Pro behaves on the smaller model, so if you have built any, run them this week and keep the prompts somewhere outside the app.
Effort Levels and the Limit They Spend
The quieter part of the same change is an effort setting on every model you can still reach: low, medium and high. Google’s wording is that it will “increase the model’s ability to complete tasks and provide more thorough answers to queries, but will also use more of your limit”.
That limit is compute-based and it refreshes on a clock. Google’s support page says your limit “refreshes every 5 hours until you reach your weekly limit”, and sets the plans against each other as multipliers: AI Plus at twice standard limits, AI Pro at four times, Ultra at five or twenty times AI Pro depending on the subscription.
Google publishes multipliers, not counts. There is no number of prompts per day on any tier, so you cannot work out in advance what high effort costs you or when you will run out. You find out by hitting it.
What It Costs to Keep Flash and Pro
The prices, as reported by Engadget on 6 October: AI Plus $4.99 a month, AI Pro $19.99, Ultra from $99.99. The decision that matters is between free and $4.99, and $4.99 buys back Flash, not Pro.
If Pro is what you were using, the comparison is $19.99 against what you already pay elsewhere. Anyone running ChatGPT Plus or Claude Pro already has a comparable reasoning model on a plan they are paying for, and a second subscription at the same price for the same job is hard to justify. If Gemini is your only paid AI tool, $19.99 is the number. If it is your third, this is a good week to cut it and compete with bigger brands on one stack you know well rather than three you half use.

What to Check Before 9 October
- Open the model picker. Count the models on the account you actually work in, so you know what you are starting from.
- Find your Pro work. Go through the last two weeks of chats and mark which ones needed Pro and which ones were fine on anything.
- Run your skills now. If you built Gems that became skills, test each one this week while Pro is still there, and note the output you expect.
- Export the prompts. Keep anything you rely on in a document outside the Gemini app, so a model change cannot strand it.
- Test Flash-Lite deliberately. Switch to it on purpose for a day and send it your real work, not a sample question.
- Check your other subscriptions. Write down what you already pay for and what reasoning model each one includes before adding another.
- Watch for the email. AI Plus subscribers get their own effective date by email, so the change may land for you after 9 October.
Questions About the Gemini Free Tier
Will my saved Gems and skills stop working?
Google has not said they will stop working, and skills do not require a paid plan. What changes is the model underneath them on a free account, which becomes Flash-Lite only. Treat a skill that depended on Pro-level reasoning as untested rather than broken, and check it yourself before the date.
Can I buy access to Gemini 4 Argon?
Not as a consumer, not yet. Google says broader release starts with paid API customers and AI Ultra subscribers, and has given no date for either. Argon’s first users are cyber defenders in the Fairwind Program.
Does this affect a work or school Google account?
Google’s notice covers the consumer plans: free, AI Plus, AI Pro and AI Ultra. Nothing has been published about model access on work or school accounts, so if your agency runs on Workspace, verify it in your own admin console rather than assuming the consumer change applies.
Is Flash-Lite good enough to keep working for free?
For high-volume drafting, summarising and rewriting, yes. For multi-step analysis and long comparisons, it is a real step down from Pro. The test that settles it is sending Flash-Lite your own recurring tasks for a day rather than reading anyone’s benchmark.
The move this week is not choosing a plan. It is finding out which of your tasks genuinely needed Pro, because that list is shorter than most people expect, and it decides whether you owe Google $4.99, $19.99 or nothing.
This is the kind of thing we work on with clients at Social Lady. If you want a second opinion on which AI subscriptions are actually earning their place, get in touch.





