October 4, 2026 · Daily Radar
Five days. Starting October 9, everyone using Gemini with a free account will see one model in the app instead of three: Flash-Lite. Google's smallest, fastest, but shallowest model.
The people who should worry about this are the ones who brought Gemini into their work because "it's free anyway." The accountant writing proposal copy, the e-commerce owner producing product descriptions, the clinic receptionist tidying up customer emails in Gemini, the café having it write Instagram captions. Most of them work with a personal Gmail account. When I read this news, the first thing I did was look at my own list: which job runs on which account?
Quick answer: starting October 9, 2026, Google is limiting the Gemini app on personal accounts without a subscription to the Flash-Lite model only; Pro is also going away for AI Plus subscribers, with the date to be announced by email. Freelancers and small businesses running their work on free Gemini are directly affected. What to do: this week, try your three most frequent Gemini tasks on Flash-Lite and see whether the result still works for you.
A page titled "Changes to Gemini model access and limits" was updated in Google's Gemini help center. The page says model access on personal accounts will change starting in October 2026, and gives the date: October 9 for users without a subscription. [A]
Here's the picture after the change:
There are two more notes on the same page. You'll be able to choose a low, medium, or high "effort" level for each model; high effort gives more thorough answers but eats more of your limit. And it recalls the May change: since May 17, Gemini has set limits based on processing load rather than message count, with the limit refreshing every 5 hours and a weekly cap on top. [A]
9to5Google was the first to report the news (October 3, 1:57 p.m. Pacific time). They also add the model versions: Flash-Lite 3.5, Flash 3.6, Pro 3.1. Google's page doesn't give version numbers, only the family names. [B]
Because the free tier isn't just losing models; it's staying inside a box that was already narrow.
According to Google's limits page, an account without a plan has a context window of 32 thousand tokens; it's 128 thousand on AI Plus, and 1 million on Pro and Ultra. [A] So when you upload a 40-page contract or a year of customer correspondence to a free account, the model can't "read" the whole document at once. Add the smallest model on top of that, and the chance of missing details in a long document goes up. That last sentence is my inference; Google hasn't published any such measurement.
In Turkey: pricing is a separate issue. According to a Technopat report from August 9, Google One and AI plans in Turkey went up by about 75% starting July 27. [C] The plan contents and lira amounts don't fully match across sources, so I'm not putting numbers here. Check your own account's current price on the Google One screen.
Then there's the timing. 9to5Google writes that Google announced its new frontier model, Gemini 4 Argon, this week and will open it to Ultra subscribers first. [B] Narrowing the free tier the same week a new model arrives doesn't look like a coincidence. But Google hasn't explained the reasoning, and I don't know whether it's a cost calculation, a sales funnel, or both.
I think the real issue isn't the model name. The real issue is that in many small businesses, AI use started in a way nobody decided on. Someone opened Gemini one day, it was useful, and then that workflow settled in there. Nobody asked "which account, which model, and which data is this running on?" October 9 is going to ask you that question.
| Before October 9 | After October 9 | |
|---|---|---|
| Free account, available models | Flash-Lite, Flash, Pro | Flash-Lite only |
| Free account, context window | 32 thousand tokens | 32 thousand tokens (unchanged) |
| AI Plus, Pro model | Yes | Going away (date by email) |
| AI Pro, Deep Think | No | Yes |
| How limits work | Based on processing load, refreshing every 5 hours | Same, plus an effort level per model |
Pick the three most important jobs you've had Gemini do in the past week. Run the same prompts again today by switching to Flash-Lite in Gemini's model picker, and put the results side by side with the old output. If there's no difference, do nothing. If there is, make a decision for that job: a paid plan, a different tool, or taking the work back from AI? None of these costs money, and it takes half an hour.
| News | Why people are talking about it | Who it concerns |
|---|---|---|
| Adjust's Turkey report: e-commerce app sessions grew 4% year over year in the first half of 2026, versus 0.3% globally (September 30) [B] | Rare Turkey-specific app data; finance apps retain 4% of users on day 30, twice the global rate | E-commerce brands with a mobile app, app advertisers. Google Ads |
The st_source=ai_overview parameter seen on AI Overviews links turned out to be the work of a Chrome extension, not Google (October 2) [C] | The excitement of "finally, AI traffic can be measured" fizzled the same day | People trying to separate AI traffic in GA4. GEO agency |
| Google Ads business name policy changes on October 16: advertisers with a verified relationship will be able to use a brand name different from their domain [B] | Dealers and resellers using the brand name stays explicitly prohibited | Brands working with more than one domain. Google Ads |
| Loyalty Customer Match moved to AI Mode and Gemini, with the country count rising to 14 (October 2) [B] | Member-only prices for loyalty members will show up in AI search too | Turkey isn't on the list; something to watch for now. SEO management |
| Meta's Muse agent passed 5 million downloads in the US and Canada in 22 days; Meta set aside up to half of its own ad space for Muse (Sensor Tower, September 30) [B] | Shows how much reach Meta can steer toward a new product; not yet available in Turkey | Meta advertisers, in the long run. Meta Ads |
No. The app stays free; only the model you can use drops to Flash-Lite. Limits also continue to be calculated by processing load, as they have been since May.
Google's announcement page only talks about personal accounts. There's no change for work and school accounts on this page. Your admin should check Gemini access on your Workspace plan separately.
It might be if you don't need the Pro model: Flash stays on Plus, and the context window is 128 thousand tokens. But if you do long documents, code, or multi-step analysis, the Pro model stays only on AI Pro and Ultra.
If you want to lay out, once and for all, which jobs your team uses AI for and on which accounts, take a look at the enterprise AI consulting page. If you want to set up repetitive tasks without being tied to a single account, AI automation may also help. If you'd like to see where you stand first:
When you train employees on AI, giving everyone the same seminar doesn't work. We cover three role-based levels, the modules, a time plan, data rules and how to measure whether the training worked.
AI projects often either never get started or get stuck in the pilot. Here is a 90-day plan split into three phases, discovery, pilot and rollout, with the deliverables you should have in hand at the end of each one.
According to Comscore's September 22, 2026 report, ChatGPT's share of AI assistant conversations fell from 70% to 50% in six months; the same day, Claude Opus 5.5 launched at a lower price. The risk of an AI setup tied to a single vendor, and what to do today.
We use cookies to improve the site and measure advertising performance. Analytics and advertising cookies run only if you accept them. Details are in our Privacy Policy and Data Protection Notice.
