Google Launches Gemini 2.5 Pro, Its Most Advanced AI Model to Date

​​​​​​​In the view of Google, the new models aim to examine information, draw logical conclusions, account for context and subtlety, and come to well-reasoned conclusions.

Google has announced its newest AI models within the Gemini 2.5 series, starting with the trial Gemini 2.5 Pro variant. In a recent posting on the Google DeepMind blog, the tech company described such models as "thinking models" and highlighted their high-level reasoning ability.

In the view of Google, the new models aim to examine information, draw logical conclusions, account for context and subtlety, and come to well-reasoned conclusions.

Advertisement

Based on the Gemini 2.0 Flash Thinking model, which was launched in December last year, the Gemini 2.5 series comes with an improved base model with improved post-training enhancements.

Google also announced its vision to bring "thinking" capabilities to all future models so that they can solve more complex problems and enhance context-aware AI agents.

Advertisement

The initial release of the new lineup is the Gemini 2.5 Pro model. Its beta version is currently rolling out to Gemini Advanced subscribers via the Gemini app and is available as well through Google AI Studio.

Referenced as Google's most advanced AI for difficult tasks, the Gemini 2.5 Pro Experimental model exhibits superb reasoning and coding skills. As per the firm, it surpasses in mathematics and science testing, such as GPQA and AIME 2025. It also rates 18.8% across models without the aid of tools on Humanity's Last Exam, a test evaluating human-level understanding and reasoning.

Advertisement

As far as coding is concerned, the Gemini 2.5 Pro model is specifically efficient at creating visually appealing web applications, building agentic code solutions, and allowing for code transformation and editing. Google unveiled a preview video demonstrating the model's potential to build a fully working video game by creating executable code based on a single-line prompt.

In testing, Business Standard reported that the Gemini 2.5 Pro Experimental model has a "Show thinking" area, similar to the earlier Gemini 2.0 Flash Thinking model. The feature is transparent in that it lays out the logical process the AI uses before responding.

Advertisement

Read also| Samsung Unveils AI-Powered Galaxy Book5 Series in India

Read also| Apple’s Foldable iPhone Set for 2026 Launch, Will Feature iPhone 17 Air Technologies: Mark Gurman

Advertisement

tags
Advertisement