Abstract:
Just days after the high-profile release of the new generation flagship model Gemini 4 Argon, Google announced that it would adjust the Gemini personal account access policy and significantly reduce the range of models available to free users. The new policy will officially take effect on October 9, 2026. At that time, users who have not subscribed to any Google AI paid plan will only be able to access Flash-Lite models and lose access to Gemini Flash and Gemini Pro models.
According to Google’s updated support document, the three model combinations of Flash-Lite, Flash and Pro currently available to free users will be cancelled. After the adjustment is completed, the free account will only retain Flash-Lite access qualifications, which means that the capabilities and usage scenarios of the free version of Gemini will be significantly limited.
Paid users will also be affected to a certain extent. Google stated that Google AI Plus subscribers will only be able to use Flash-Lite and Flash models in the future, and the originally accessible Pro models will be moved to higher-level AI Pro and AI Ultra subscription plans. Google will notify Plus subscribers via email of the specific effective time.
Under the new subscription system, AI Pro and AI Ultra will still retain access to Flash-Lite, Flash and all Pro models. In addition, these two high-end subscription plans will also receive Deep Think feature support. This feature was previously only available to AI Ultra users and was mainly used to perform more complex parallel reasoning tasks.
The adjusted model permission structure shows that the Gemini product system is further transforming into a tiered charging model. Free users can only use Flash-Lite, AI Plus users can use Flash-Lite and Flash, and AI Pro and AI Ultra users can get a complete model portfolio and advanced inference capabilities.
This adjustment coincides with the release of Gemini 4 Argon by Google. Argon is positioned as Google's new generation flagship artificial intelligence model, benchmarking OpenAI's GPT-6 Astra and Anthropic's Claude Opus 5.5 in multiple professional fields such as software engineering, legal services, financial analysis, and network security. Data previously released by Google shows that Argon has achieved leading results in multiple benchmark tests and supports ultra-long output capabilities of up to 1 million Tokens.
However, Gemini 4 Argon is not currently open to ordinary users, but will be given priority to network security organizations and some partners participating in the Google Fairwind program for testing. It was originally expected that Google would use Argon to further expand the user base of Gemini, but this permission adjustment means that the model capabilities accessible to ordinary free users are decreasing.
At the same time, Google has also enhanced functions for high-end subscribers. As the Deep Think function is decentralized to the AI Pro level, premium subscriber users can gain stronger complex task processing and reasoning capabilities, thus forming a more obvious distinction from free and basic paid users.
Industry insiders believe that this adjustment reflects the reality that the operating costs of large-scale artificial intelligence models continue to rise. As model capabilities continue to increase and inference resource consumption increases simultaneously, major AI vendors are gradually shrinking the scope of free services and concentrating the most advanced capabilities in high-level subscription plans. For Gemini users, the range of models available after October 9 will undergo the most significant change in recent years.
Comments