Particle.news

Download on the App Store

Google Launches Gemini 2.5 Pro, Its Most Advanced AI Model Yet

The multimodal reasoning model sets new benchmarks in AI performance and is now available to Gemini Advanced users and developers.

Gemini 2.5 is Google's new AI model.
Image
Google
Image

Overview

  • Gemini 2.5 Pro Experimental introduces advanced reasoning capabilities, outperforming competitors on benchmarks like Humanity's Last Exam with a state-of-the-art score of 18.8%.
  • The model features a 1 million token context window, enabling it to process vast datasets, with plans to expand this to 2 million tokens soon.
  • Gemini 2.5 Pro excels in coding tasks, achieving 68.6% on Aider Polyglot and 63.8% on SWE-Bench Verified, surpassing most rivals in these benchmarks.
  • It is natively multimodal, capable of processing text, audio, images, video, and code, making it versatile for enterprise applications.
  • Available now in Google AI Studio and the Gemini app for Advanced users, the model will integrate with Vertex AI in the coming weeks, with pricing details to follow.