Gemini 2.5: Here’s What Google’s ‘Most Intelligent AI Model’ Is Capable Of

Google has introduced Gemini 2.5, calling the artificial intelligence model its most intelligent. The first 2.5 release is an experimental version of 2.5 Pro, which has debuted at No. 1 on LMArena benchmarks by a significant margin.

Google said Gemini 2.5 models are thinking models, and they are capable of reasoning and analysing their responses before acting, which improves accuracy and performance. This means these models can analyse information, draw logical conclusions, incorporate context and nuance, and make informed decisions.

Gemini 2.5 Pro Experimental is Google’s “most advanced model for complex tasks,” having reasoning and code capabilities, leading on common coding, math and science benchmarks. Here’s a look at Gemini 2.5 Pro’s capabilities:

Enhanced Reasoning: Gemini 2.5 Pro leads on a variety of benchmarks that need advanced reasoning: 2.5 Pro leads in math and science benchmarks like GPQA and AIME 2025. It also scored 18.8% across models without tool use on Humanity’s Last Exam, a dataset designed by subject matter experts to capture the human frontier of knowledge and reasoning.

Advanced Coding: Google said Gemini 2.5 makes significant progress over 2.0 in coding performance. 2.5 Pro is proficient in transforming and editing code, as well as developing web apps and agentic code applications. With a customised agent configuration, Gemini 2.5 Pro achieves 63.8% on SWE-Bench Verified, the industry benchmark for agentic code evaluations. 

The native multimodality and extended context window that distinguish Gemini models are enhanced in Gemini 2.5, Google said. 2.5 Pro has good performance that surpasses earlier versions. Large datasets and complicated issues from various information sources, including text, audio, images, video, and even complete code repositories, can be understood and handled by it.

Google is launching Gemini 2.5 with a 1 million token context window (2 million in the pipeline). Gemini Advanced users may choose it from the model dropdown on desktop and mobile devices, while developers and businesses can begin with Gemini 2.5 Pro in Google AI Studio. In the upcoming weeks, it will also be available on Vertex AI.