AI-generated summary
Starting in May, Google is changing Gemini usage limits to be calculated based on factors such as the complexity of the prompt, the features used, and the length of the chat, rather than a simple number of times.
On the Gemini support page that was updated in early October, Google announced that the AI models that can be used in the personal Gemini app will change depending on the Google AI contract plan that you subscribe to. For free users who do not subscribe to a Google AI paid plan, the change will be applied from October 9th, and the only model available will be "Flash-Lite."
After the change, there will be three models available: "Flash-Lite," "Flash," and "Pro." Free users can only use Flash-Lite, "Google AI Plus" plans can use Flash-Lite and Flash, and "Google AI Pro" and "Google AI Ultra" can use all three models. The subscription fees for each plan are as follows.
In other words, the free version of Gemini itself is not unusable, but the choice of available models is limited. To use Flash, you will need an AI Plus or higher contract, and to use the top-of-the-line Pro model, you will need an AI Pro or higher contract.
Each model has a specification that allows you to select the amount of processing allocated to inference from ``low'', ``medium'', and ``high''. The higher the setting, the better the ability to handle complex tasks and the detail of answers, but the higher the usage limit. AI Pro and AI Ultra also offer Deep Think, which uses parallel reasoning to tackle more complex problems.
Starting in May, Google is changing Gemini usage limits to be calculated based on factors such as the complexity of the prompt, the features used, and the length of the chat, rather than a simple number of times. The limit is updated every 5 hours, and there is also a weekly limit. If we consider the free version to be "standard," AI Plus has double the usage limit, and AI Pro has four times the usage limit. AI Ultra is 5 times or 20 times more expensive than AI Pro depending on the contract details. Image/video/music generation, Deep Research, Pro models, extended inference, etc. consume a relatively large amount of usage quota.
This model change applies to personal accounts. It is scheduled to start applying to free users from October 9th. AI Plus users will be notified via email when the changes will take effect.
AI outlook — possibilities, not facts
Some free users will migrate to paid plans, increasing Google's AI-related revenue
Likely · Within months

The Australian government is considering regulating the wearing of smart glasses in public facilities due to concerns about security risks, such as the possibility of voyeurism and the leaking of confidential information. Minister for Public Services Gallagher has instructed that a ban be considered.

NTT Docomo announced that RCS was available to approximately 16,000 users who signed up for new MNP contracts at stores between September 17th and September 25th without receiving an explanation of important matters regarding the use of RCS or obtaining consent for third-party provision to Google Asia Pacific. There were 225 messages that were actually sent and received, and the messages were encrypted, but this was considered problematic from the perspective of secrecy.
David Robinson, who was in charge of safety reports for major models at OpenAI, has left the company and criticized the company's culture as broken in an article he wrote for The Atlantic. He pointed out the accumulation of risks due to iterative development methods and called for a culture of safety at the level of nuclear power plants and airports. Similar concerns have been raised by other AI companies.

Gartner positions frontier AI, FDE, agent-based AI, physical AI, and cloud resilience as five items in the peak period, and analyzes frontier AI as a core technology of the AI industrial revolution and a strategic foundation that determines national technological sovereignty. He emphasized that the cloud has evolved to be AI-driven, and that the cloud strategy is also a human resources strategy. We recommend that companies strengthen their competitiveness through comprehensive AI strategies, including multi-AI model strategies and sovereign AI, and the continuous evolution of cloud infrastructure.

On September 30th, Google announced Gemini 4 Argon, a new frontier model with enhanced long text processing and deep reasoning. The output token limit has been raised to 1 million, emphasizing its performance in the software development and cybersecurity fields. OpenAI released “GPT-6.1 Sol” on September 29th, in time for DevDay 2026. Claiming improved performance in agent-based coding and business workflows, the API fee is $2 for input and $10 for output (per 1 million tokens), with a 95% discount on cached input. Both companies have strengthened safety measures and started offering the service in stages.
Matt Garman, CEO of Amazon's AWS division, published an official blog on October 2 to refute criticism in the United States regarding the construction of data centers. They also announced a code of conduct and a new investment program called ``Built Together.''