Google has rolled out a preview of its upgraded Gemini 2.5 Pro, calling it the company’s most advanced AI model to date.
Building on the foundation laid at its I/O developer conference in May, the new version significantly raises the bar in performance across industry-standard benchmarks. According to Tulsee Doshi, Senior Director of Product Management at Google, the model soared by 24 points on LMArena, reaching a leading score of 1470. On WebDevArena, which measures web development proficiency, it improved by 35 points to top the chart at 1443.
Gemini 2.5 Pro also excelled in the Aider Polyglot test, showcasing its strength in complex, multilingual coding. Additionally, it placed among the best performers in GPQA and Humanity’s Last Exam, which evaluate higher-order reasoning in math, science, and general knowledge.
Developers can now start testing the model using Google AI Studio and Vertex AI, while the enhanced model is also accessible through the Gemini app. General availability is expected in the coming weeks.
Google noted that improvements were made based on developer feedback, highlighting the company’s commitment to refining its AI models through community engagement. The early release follows high demand and enthusiasm from the developer community.
With the upgraded Gemini 2.5 Pro, Google continues its strategic push in the competitive AI race, especially as OpenAI’s ChatGPT remains dominant in usage. The tech giant is placing a strong emphasis on innovation in areas such as code generation and scientific reasoning.
Originally introduced in March, Gemini 2.5 is part of Google’s flagship AI family. It stands out as the first major AI model built from the ground up to handle multimodal inputs — including text, images, code, and video — with advanced reasoning capabilities that go beyond simple classification.
Want more marketing excellence stories? Visit MarTech News for insights, trends, and expert updates.
News Source: Pymnts.com
