Gemini model access is about to shrink for anyone who does not pay Google. From Thursday 9 October 2026, people using the Gemini app on a personal Google account without a Google AI plan will be limited to Flash-Lite, the smallest of Google’s three Gemini models. Flash and Pro move behind a subscription, and the cheapest paid plan, Google AI Plus, loses Pro as well.
Google published the change in a help article titled “Changes to Gemini model access and limits”, and 9to5Google first reported it on 3 October. The same weekend brought the opposite kind of news from Germany. Aleph Alpha released Kolibri-1, an open-weight model whose full weights anyone can download, run and keep under the Apache 2.0 licence.
The two stories make a useful pair. One shows a frontier lab rationing its best models to people who pay; the other shows what owning a model outright looks like. Below we set out exactly what changes to Gemini model access on 9 October, what free and paying users gain or lose, with UK prices, why Google is doing it, where Kolibri-1 fits, and what UK businesses should check before Thursday.
Table of contents
- What Changes to Gemini Model Access on 9 October
- How Gemini Model Access Worked Before
- What Free Users Lose Under the New Gemini Model Access Rules
- What Paying Subscribers Gain From the Gemini Model Access Change
- Why Google Is Tightening Gemini Model Access
- The Other Option: Aleph Alpha’s Open-Weight Kolibri-1
- How to Check and Manage Your Gemini Model Access
- What the Gemini Model Access Change Means for UK Businesses
- Gemini Model Access FAQ
- References and Further Reading
What Changes to Gemini Model Access on 9 October
The Gemini model access change is short to state and has several details that the headlines skip.
The new Gemini model access table
Google’s help article lists which models each plan will get once the change takes effect. We have added Google’s current UK and US monthly prices from its subscriptions pages.
| Plan | UK price a month | US price a month | Flash-Lite | Flash | Pro | Deep Think |
|---|---|---|---|---|---|---|
| No plan | £0 | $0 | Yes | No | No | No |
| Google AI Plus | £4.49 | $4.99 | Yes | Yes | No | No |
| Google AI Pro | £18.99 | $19.99 | Yes | Yes | Yes | Yes |
| Google AI Ultra | £79.99 or £189.99 | $99.99 or $199.99 | Yes | Yes | Yes | Yes |
Under the new Gemini model access rules, AI Pro is the cheapest plan that includes all three models. The two Ultra prices buy different usage allowances: five times AI Pro’s limit at the lower price and twenty times at the higher one.
Who loses Gemini model access, and when
The Gemini model access change applies to Gemini Apps on personal Google accounts. Users without a plan move on 9 October. AI Plus subscribers “should receive an email that explains when these changes will take effect”, so their date may differ. The help article does not mention work or school accounts, which suggests Google Workspace customers are outside this round.
Users are also being told in the app. OfficeChai reported a pop-up saying Flash and Pro will soon move to paid plans, “while unlimited access to the latest Flash-Lite will continue”.
What Google has not said
Three gaps matter. Google does not name the Flash-Lite version free users will get; 9to5Google lists it as Gemini 3.5 Flash-Lite. Google does not say whether existing chats started on Flash or Pro will switch to Flash-Lite. And, as Notebookcheck noticed, Google’s US plans page still listed “Access to 3.6 Flash” and “Varying access to 3.1 Pro” for the free tier after the help article appeared. Only the new article describes the post-9 October Gemini model access.
How Gemini Model Access Worked Before
The new Gemini model access restriction is the second big change to the app this year. The first one changed how usage is counted.
Compute-based limits since May
Since 17 May 2026, the Gemini app has used compute-based limits rather than fixed prompt counts. Users under 18 moved on 24 July. Your allowance refreshes every five hours until you hit a weekly cap, and usage is calculated from the complexity of the prompt, the features used and the length of the chat.
Premium models and features use the allowance faster. Google names media generation, Deep Research, the Pro model, and extended thinking or Deep Think. Paid plans get multiples of the standard limit: AI Plus twice, AI Pro four times, and AI Ultra five or twenty times the AI Pro level.
Context windows by plan
The plans also differ in how much text Gemini can read at once. Google’s limits page gives 32,000 tokens without a plan, 128,000 tokens on AI Plus and 1 million tokens on AI Pro and AI Ultra. Google says a 1 million token window can hold about 1,500 pages of text or 30,000 lines of code.
Gemini app context window by plan (thousands of tokens)
Bars are scaled to the 1 million token window. AI Pro reads 31 times as much as the free tier (1,000 ÷ 32 is 31.25) and about 8 times as much as AI Plus (1,000 ÷ 128 is 7.8). Using Google’s own estimate of 1,500 pages per million tokens, the free tier’s 32,000 tokens comes to roughly 48 pages (1,500 × 0.032).
What Free Users Lose Under the New Gemini Model Access Rules
For most casual users, the switch may go unnoticed. For anyone who leaned on the free tier for real work, it is a meaningful cut.
From Flash to Flash-Lite
Google describes Flash-Lite as “an efficient workhorse model designed for speed”, ideal for “everyday tasks like summarization and brainstorming”. Flash, which free users lose, “balances speed and reasoning to solve a large variety of problems, from simple to complex”. The drop is from Google’s mid-tier model, which has received most of its updates this year, to its smallest one.
That timing stings. Google released Gemini 3.8 Flash on 2 September, after Gemini 3.6 Flash and 3.5 Flash-Lite in July, and Sundar Pichai described Gemini 3.7 Flash as Google’s fastest-growing model ever, according to OfficeChai. The model family Google has promoted hardest now sits behind a paywall for the free app.
No more occasional Pro
Free users currently get “varying access” to Gemini 3.1 Pro. That Gemini model access ends on 9 October. Notebookcheck’s advice is practical: if you have used Pro without a subscription for long documents, code or difficult maths, finish that work before the cut-off.
What stays free
Unlimited access to the latest Flash-Lite continues, within the compute-based limits. Google also says that subscribers who hit their limit can keep chatting on Flash-Lite, which makes Flash-Lite the floor for everyone, paid or not.
What Paying Subscribers Gain From the Gemini Model Access Change
The Gemini model access update is not only a cut. Paying users get two new options, and one paid plan loses something.
Effort levels for every model
For each available model, users will be able to choose a low, medium or high effort level. Google says higher effort improves the model’s ability to complete tasks and gives more thorough answers, “but will also use more of your limit”. Today the app offers only an on-off “extended thinking” option. 9to5Google notes the three levels match those already in Google AI Studio and Antigravity.
Deep Think reaches AI Pro
Deep Think, which Google describes as “maximum parallel reasoning”, is currently limited to AI Ultra. After the change, AI Pro subscribers get it too. Google warns that Deep Think answers can take a few minutes, and it needs the Pro model.
AI Plus loses Pro
The £4.49 AI Plus plan keeps Flash-Lite and Flash but loses Pro. That widens the gap between the two cheaper paid plans, and it makes AI Pro the default recommendation for anyone who needs Google’s strongest model.
Google AI plan prices in the UK (£ a month)
Bars are scaled to the £189.99 plan. AI Pro costs about 4.2 times AI Plus (18.99 ÷ 4.49), and the top Ultra plan costs 10 times AI Pro (189.99 ÷ 18.99). Over a year, AI Pro comes to £227.88 (18.99 × 12), and AI Plus to £53.88 (4.49 × 12).
Why Google Is Tightening Gemini Model Access
Google has not explained the new Gemini model access rules beyond the help article. The coverage points to three.
Serving costs
Flash and Pro are the models that use the most compute. Moving them off the free tier caps Google’s biggest variable cost. The Decoder’s Matthias Bastian argued that the real-world impact is “likely small”, because most casual users do not know which model they are running and power users are probably paying already. He added that the new structure “mirrors OpenAI’s freemium model with ChatGPT”.
Making room for Gemini 4 Argon
Google introduced Gemini 4 Argon to partners last week and said it would come first to AI Ultra subscribers. We covered its early tests in our Gemini 4 Argon report. The Decoder suggests the new tiers may set the stage for a model that is more expensive to run. Google has not said whether Argon will count as a Pro model or sit in a tier of its own.
Everyone is rationing
Google is not alone. Notebookcheck noted that Anthropic kept Claude Opus 5.5 for paying customers in September. Usage caps, effort settings and paid-only frontier models are now standard across consumer AI, as our coverage of ChatGPT’s five-hour usage limit showed. The era of a free flagship model is ending.
The Other Option: Aleph Alpha's Open-Weight Kolibri-1
While Google narrowed free Gemini model access, Heidelberg’s Aleph Alpha went the other way. Kolibri-1 is a model you download once and keep.
What Kolibri-1 is
Released on 3 October 2026, Kolibri-1 is an English and German mixture-of-experts model with 78.1 billion parameters, of which 3.46 billion are active for each token. It handles German and English natural language processing with reasoning levels and tool calling, and its weights are on Hugging Face under the Apache 2.0 licence, which allows commercial use. Aleph Alpha pitches it at regulated work in public administration, industry and aerospace. We covered the model in depth in our Aleph Alpha Kolibri analysis.
The 1M context footnote
Headlines say Kolibri-1 has a 1 million token context window. It was trained on sequences of up to 262,144 tokens, and the jump to a million relies on a design that lets it run past its training length. Aleph Alpha’s own model card recommends staying at or below 262,144 tokens for serving efficiency and complex tasks. In its long-context tests, the base model scores 86.9 at 4,000 tokens and 63.2 at a million.
What it takes to run
The default FP8 download is about 78 GB, and all 78 billion parameters must sit in GPU memory even though only 3.46 billion are active per token. Aleph Alpha’s minimum is two A100 80 GB cards, two H100s, or a single H200, B200 or B300. These are data-centre GPUs, not desktop cards, so most firms would rent them or use a hosting partner.
Rented access or owned weights
The contrast with Gemini model access is about control, not just price. A subscription can change on a week’s notice, as this one has. Open weights cannot be withdrawn once downloaded, but you carry the hardware, security and maintenance.
| Question | Free Gemini app | Google AI Pro | Kolibri-1, self-hosted |
|---|---|---|---|
| Monthly cost | £0 | £18.99 per user | GPU hire or hardware |
| Best model available | Flash-Lite | Pro, with Deep Think | Kolibri-1 |
| Context window | 32,000 tokens | 1 million tokens | 262,144 recommended |
| Can terms change? | Yes, at Google’s discretion | Yes, at Google’s discretion | No, once downloaded |
| Where data goes | Wherever you host it | ||
| Languages | Many | Many | English and German |
Kolibri-1 is not a like-for-like replacement for the Gemini app. It is a model, not a finished assistant, and it covers two languages. But for a firm that needs predictable terms or must keep documents in-house, it is a credible option. Aleph Alpha is also being folded into Canada’s Cohere under the Cohere and Aleph Alpha merger, which signed a definitive agreement on 16 September; the Apache 2.0 weights already released are not affected by that.
How to Check and Manage Your Gemini Model Access
Google’s help pages explain how to see where your Gemini model access stands before and after the change, and what happens when you hit a limit.
Find your current usage
On the web, open gemini.google.com, select Settings at the bottom left, then Usage Limits. Gemini also warns you when you are close to your limit and tells you when it will refresh. Because limits are compute-based, one long chat with a large file can use more of your allowance than several short questions.
Choose the right effort level
Once the effort setting arrives, low effort will be the economical choice for quick questions, and high effort the choice for tasks where a more thorough answer is worth spending more of your limit. On the free tier, Gemini model access is limited to Flash-Lite, so every effort choice applies to that one model. On paid plans, picking Flash at medium effort instead of Pro at high effort will stretch the same weekly allowance further.
Know what happens when you run out
Subscribers who reach their limit can carry on with Flash-Lite. Users without a plan who reach a five-hour or weekly limit can wait for the refresh or upgrade. Google also warns that, for users without a plan, features that need more compute, such as Deep Research, may be unavailable at busy times, and that free users’ limits may be cut first if capacity is tight.
Upgrade, downgrade or cancel
Google says you can subscribe, upgrade, change or cancel a Google AI plan at any time from within the Gemini app. If you only need Pro for a short project, one month of AI Pro at £18.99 is the cheapest way to get it, because AI Plus no longer includes Pro at any price. Cancelling afterwards returns you to the free tier’s Gemini model access, with Flash-Lite only.
What the Gemini Model Access Change Means for UK Businesses
Most firms that use Gemini at work do so through Google Workspace, which this change does not mention. The risk sits in the gaps.
Check which accounts your staff actually use
Staff often use personal Google accounts for work tasks, especially in small firms. Those people lose Flash and Pro on 9 October. If someone’s workflow depends on the free tier’s Pro access, it will break quietly. This is also a good moment to ask whether company documents should be going into personal AI accounts at all.
Budget for the right tier
For someone who needs Pro and long documents, AI Pro at £18.99 a month is now the entry point for full Gemini model access, at £227.88 a year per person. AI Plus at £4.49 suits lighter users who only need Flash. Compare those with your Workspace plan’s Gemini features before buying personal subscriptions.
Match the model to the data
Sensitive documents belong in tools your organisation controls. For some UK firms, especially those working with German-language material or in regulated sectors, an open-weight model such as Kolibri-1, hosted in a UK or EU data centre, may fit better than any consumer subscription. Our guide to open-weight AI models compares the main options.
A checklist before Thursday
| Check | Why it matters | Action |
|---|---|---|
| Who uses personal Gemini for work | They lose Flash and Pro on 9 October | Move them to a managed account |
| Unfinished Pro work on free accounts | Chats may switch to Flash-Lite | Finish or export before Thursday |
| Long documents on free accounts | 32,000 tokens is about 48 pages | Use AI Pro or a managed tool |
| AI Plus subscribers who need Pro | Plus loses Pro | Watch for Google’s email; upgrade if needed |
| Sensitive data in consumer AI | Data leaves your control | Review policy; consider self-hosted models |
Gemini Model Access FAQ
When does the Gemini model access change happen?
On 9 October 2026 for users without a Google AI plan. AI Plus subscribers will get an email from Google telling them when the change applies to them.
Which Gemini models can free users use after 9 October?
Only Flash-Lite. Flash and Pro require a paid plan. Google does not name the version; 9to5Google lists it as Gemini 3.5 Flash-Lite.
Does AI Plus still include Gemini Pro?
No. AI Plus, £4.49 a month in the UK, keeps Flash-Lite and Flash but loses Pro. AI Pro, at £18.99, is now the cheapest plan with all three models.
Does the change affect Google Workspace accounts?
Google’s help article covers Gemini Apps on personal accounts and does not mention work or school accounts.
What is Deep Think, and who gets it?
Deep Think is Google’s “maximum parallel reasoning” option for the Pro model. It is currently limited to AI Ultra and will also be available on AI Pro.
Is Kolibri-1 an alternative to Gemini?
It is a different kind of product. Kolibri-1 is an open-weight English and German model you host yourself under Apache 2.0. It suits firms that want fixed terms and in-house data, but it needs data-centre GPUs and is not a ready-made assistant.
References and Further Reading
Changes to Gemini model access and limits (Gemini Apps Help)
Gemini Apps limits and upgrades for Google AI subscribers (Gemini Apps Help)
Google AI plans and prices, United Kingdom (Google)
Gemini limits which models free and AI Plus users can access (9to5Google)
Google Gemini drops Flash and Pro for free users on October 9 (Notebookcheck)
Google’s new Gemini tiers cut free users to its weakest model (The Decoder)
Gemini Flash and Gemini Pro models to no longer be available to free Gemini users (OfficeChai)
Google is downgrading Gemini for users on budget plans (Android Authority)
Kolibri-1 model card (Hugging Face)
Kolibri has landed: a sovereign open-weight model (Aleph Alpha)
Aleph Alpha releases Kolibri, a 78.1B open-weight English-German MoE model (MarkTechPost)
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.
