Google Gemini 2.5: The New AI Model with Advanced Reasoning Capabilities

Gemini 2.5 Pro Experimental positions itself as the leading model in the LMArena ranking, significantly surpassing other artificial intelligence models. Its most innovative feature is the "reasoning" capability, which allows it to analyze information, incorporate context and nuances, and make informed decisions before responding.

The new model shows outstanding performance in various fields. In math and science benchmarks like GPQA and AIME 2025, Gemini 2.5 Pro leads the results. It also achieves a notable 18.8% in Humanity's Last Exam, a dataset designed to evaluate the limits of human knowledge and reasoning.

In programming, Gemini 2.5 Pro especially excels in web development, code generation, and complex programming tasks. In SWE-Bench Verified, an industry standard evaluation, it obtains a score of 63.8%, consolidating itself as a reference model in coding capabilities.

One of its most powerful features is its native multimodal nature, capable of understanding and processing information in text, audio, images, and video. It also has a context window of one million tokens, which will soon be expanded to two million, allowing exploration and analysis of extensive datasets.

Google has already made this new model available to developers and Gemini Advanced users through Google AI Studio. In the coming weeks, it will also be available in Vertex AI, with plans to introduce new pricing options for professional use.