Google Gemini app to restrict model access for free and AI Plus users

Google is cutting access to its more powerful artificial intelligence models for millions of Gemini app users, leaving non-paying accounts with only its most basic model.

Starting on 9 October, users without a subscription will lose access to both the Flash and Pro models, Google confirmed in an update to its support pages. Instead, the free tier of the app will be restricted entirely to Gemini Flash-Lite, a smaller model designed for quick, simple tasks.

The downgrade also hits paying customers on Google's entry-level subscription. Users on the AI Plus plan, which costs $4.99 a month in the US, will keep Flash-Lite and Flash but will lose access to the flagship Gemini Pro model. Google has not set a single date for this change, saying instead that it will email subscribers individually to let them know when the Pro model will disappear from their accounts.

What the changes mean for subscribers

The reshuffle leaves Google's subscription tiers starkly divided. While the lower tiers lose capability, the premium plans are getting an upgrade.

Subscribers paying $19.99 a month for the AI Pro tier will keep access to the Pro model and gain "Deep Think", a parallel reasoning feature that was previously restricted to the ultra-premium plans starting at $99.99 a month. AI Ultra subscribers also keep access to all models and Deep Think.

The timing of the cuts is notable, coming just over a week after Google announced its new Gemini 4 Argon model. While Google has not officially explained the decision, the shift is likely a move to manage server strain and computing costs. Running high-end AI models like Pro is expensive, and Google has struggled to balance the compute demands of millions of active free users against subscription revenues.

The changes are also bound to frustrate Google One users, particularly in Germany, where some cloud storage packages came bundled with AI Plus benefits — and access to the Pro model — that users did not explicitly sign up for.

Adjusting model effort levels

Alongside the tier changes, Google is introducing new manual controls that let users dictate how hard a model works on a query.

Later this month, the Gemini app will add "low", "medium", and "high" effort levels for each available model. Setting a model to high effort will prompt it to complete tasks more thoroughly and provide deeper answers, but it will also consume more of a user's compute-based limit.

These limits, which Google introduced in May, reset every five hours based on the complexity of prompts rather than a simple count of messages. The AI Plus tier gets double the standard compute limit, while AI Pro users get four times as much.

For users who rely on Gemini for complex coding, writing, or analysis, the upcoming deadline means they must either downgrade their expectations to the lightweight Flash-Lite model or prepare to pay a significantly higher monthly bill to step up to the AI Pro tier.