Google introduced Gemini 2.5 — an AI that outperforms models from OpenAI

Google has introduced Gemini 2.5, a new family of artificial intelligence models that "think" before answering questions.

To showcase the new model family, Google is launching Gemini 2.5 Pro Experimental — a multimodal AI model that the company claims is the most intelligent among all its models. This model will be available on Tuesday on the Google AI Studio developer platform, as well as in the Gemini app for subscribers to the $20-per-month Gemini Advanced plan.

Reasoning techniques have helped AI models achieve new heights in solving mathematical and programming tasks. Many in the tech industry believe that reasoning models will become a key component of AI agents — autonomous systems capable of performing tasks with minimal human intervention.

Gemini 2.5 Pro outperforms previous state-of-the-art AI models and some leading competing AI models across several metrics. In particular, Google states that Gemini 2.5 is designed to create visually appealing web applications and support agentic coding.

In the Aider Polyglot code editing benchmark, Gemini 2.5 Pro scored 68.6%, surpassing top AI models from OpenAI, Anthropic, and the Chinese lab DeepSeek.

In another test measuring software engineering capabilities, SWE-bench Verified, Gemini 2.5 Pro scored 63.8%, outperforming OpenAI's o3-mini and DeepSeek's R1, but falling short of Anthropic's Claude 3.7 Sonnet, which scored 70.3%.

In Humanity’s Last Exam — a multimodal test consisting of thousands of crowdsourced questions related to mathematics, humanities, and natural sciences — Gemini 2.5 Pro scored 18.8%, performing better than most flagship models from competitors.

Gemini 2.5 Pro comes with a context window of 1 million tokens, meaning the AI model can process approximately 750,000 words at once. This is more than the entire "Lord of the Rings" book series.

Source: TechCrunch