Google is tightening restrictions on its Gemini platform, announcing that free-tier users will soon be limited to using only the service’s least powerful AI model. According to a support document released by the company, the changes take effect on Friday, October 9.
From that date forward, anyone without a paid Google AI subscription will no longer have access to the Flash or Pro models. Instead, free users will be restricted to Flash-Lite, the lowest-tier option in the Gemini lineup.
The overhaul also impacts subscribers to the entry-level AI Plus plan, which costs $5 per month. These users will lose access to the Pro model, leaving them with only the Flash and Flash-Lite variants.
One notable exception involves the higher-tier AI Pro plan, priced at $20 per month. Subscribers to this plan will gain access to Deep Think reasoning mode, a feature previously exclusive to AI Ultra customers who pay between $100 and $200 monthly.
Google’s Gemini ecosystem is structured around five distinct access methods, each offering varying levels of capability. The three primary models available across these tiers include 3.8 Flash, designed as a generalist for tasks like summarization and image generation; 3.5 Flash-Lite, a faster but less capable option suited for repetitive tasks; and 3.1 Pro, the most analytical model intended for complex problems in math, logic, and coding.
For AI Pro subscribers, Flash typically serves as the default model for most everyday interactions. However, free users will now find Flash-Lite set as their default, which may hinder performance on tasks requiring deep reasoning or complex problem-solving.
Additionally, Google is introducing adjustable error levels for each model, allowing users to select between low, medium, and high settings. Higher error levels enable more thorough responses and greater task completion capabilities but consume compute limits more rapidly.
Usage limits are now calculated based on a compute metric rather than a fixed number of requests, refreshing every five hours until the weekly quota is reached. Factors such as prompt complexity, feature usage, and chat length determine how quickly these limits are consumed. Premium features like image generation, video creation, Deep Research, and Deep Think drain quotas significantly faster than standard text-based conversations.
While the consolidation of features and models may frustrate users accustomed to broader access, the monthly subscription structure allows consumers to upgrade or downgrade their plans flexibly depending on their needs.
Compute limits based on complexity makes sense technically, but hopefully ‘deep reasoning’ isn’t just throttled heavily.
Is anyone actually paying $200 for Ultra now? This model tiering is getting absurdly complicated.
Wait, so AI Plus subscribers lost Pro too? That feels like a bait-and-switch after the initial launch pricing.
Great, another tech giant nickel-and-diming users for basic features. Flash-Lite sounds painfully slow.