The Gemini 2.5 Pro Preview, a new AI model designed for coding and advanced reasoning, is being made available for early access ahead of its official launch at Google I/O 2025. Users can test this updated model now in Google AI Studio and Vertex AI.
This preview version boasts significant enhancements in coding performance, particularly for front-end and user interface development. Google has highlighted improvements in essential coding tasks like code transformation, editing, and the creation of intricate workflows.
It is anticipated that the full model will be unveiled at the upcoming Google I/O 2025 developer conference. Currently, Gemini 2.5 Pro Preview leads key AI benchmarks, ranking first on the WebDev Arena leaderboard, which assesses AI models based on web development performance.
It surpasses competitors, including Claude 3.7 Sonnet, placing Google at the forefront of the leaderboard with two models in the top three positions. In addition to coding capabilities, the Gemini 2.5 Pro is proficient in video comprehension, achieving an impressive score of 84.8% on the VideoMME benchmark.
Google demonstrated this feature by illustrating how a single YouTube video can be transformed into a comprehensive learning application. The Gemini 2.5 Pro Preview is accessible via the Gemini API at the same cost as the experimental version for paid users, with a free option still available for others.
This update signifies a substantial advancement in AI coding tools, making it an exciting development for developers and tech enthusiasts alike. In other news, Brady Snyder, a technology journalist at Android Central, will keep you updated on the latest in the world of Android, sharing insights on devices, software, and more.