Google introduces Gemini 2.5 Pro AI with deep thinking features

Google has introduced a new version of the Gemini 2.5 Pro AI model, which has enhanced capabilities for analyzing information and working with context.

According to the company, this model will become the basis for further development of the Gemini line and is an analogue of the o1 system from OpenAI. One of the key improvements was the so-called “reasoning”, which allows the system to analyze data more deeply, draw logical conclusions and solve complex tasks taking into account the context.

The context window of the model is 1 million tokens, and in the future it is planned to double it. In testing, Gemini 2.5 Pro showed significant progress compared to previous versions and took first place in the Humanity’s Last Exam test, where its results reached 18.8%. This test assesses the level of knowledge and logical reasoning of AI, comparing them with human capabilities.

Gemini 2.5 Pro is available for beta testing in AI Studio and Gemini Advanced subscribers. Pricing has not been revealed yet.

According to a recent APK analysis, Google is planning to expand the functionality of Gemini, its AI-powered chatbot, by adding capabilities that were previously only available through Google Assistant. Here are the main expected features:

• Alarms and Timers: Gemini will be able to set alarms and timers on its own, without the need to involve Google Assistant.
• Lock Screen: A new feature will allow users to control Gemini, reply to messages, play music, and control smart home devices right from the lock screen, without unlocking the device.
• Lock Screen Extensions: New Gemini tools, such as the Clock tool, will be available on the lock screen.
• Gemini Live: A new background mode that allows users to continue chatting in Gemini chats, even while using other apps or when the screen is locked.

Additionally, it is worth noting that some of these features may require a Gemini Premium subscription.