Gemini 2.5 Pro: Google’s New AI Model Previewed
- Google unveiled a preview of its enhanced Gemini 2.5 Pro artificial intelligence model Thursday, calling it their "most smart" model to date.
- According to Tulsee Doshi, senior director of product management at Google, the new AI model achieved notable gains on key evaluations.
- The Gemini 2.5 Pro also ranks among the best on benchmarks like GPQA and Humanity’s Last Exam, designed to test advanced reasoning, math, science, and general knowlege.
Google’s Gemini 2.5 Pro, the upgraded AI model, has just been previewed, showcasing significant performance gains across key benchmarks. This cutting-edge model, accessible now via Google AI Studio and Vertex AI, leaps ahead in reasoning and coding capabilities. News Directory 3 reports on the impressive 24-point jump on the LMArena benchmark and a 35-point surge on WebDevArena, demonstrating Gemini 2.5 Pro’s advanced abilities. Developers can dive in instantly, leveraging the enhancements to build innovative applications. With Google’s commitment to AI innovation, the future looks shining. Discover what’s next for Gemini’s evolution.
Google’s Upgraded Gemini 2.5 Pro AI Model Enters Preview
Google unveiled a preview of its enhanced Gemini 2.5 Pro artificial intelligence model Thursday, calling it their “most smart” model to date. This upgraded version of the Gemini 2.5 Pro, initially revealed in May, boasts improved performance across various industry benchmarks.
According to Tulsee Doshi, senior director of product management at Google, the new AI model achieved notable gains on key evaluations. On the LMArena benchmark, which assesses performance across a range of tasks, Gemini 2.5 Pro jumped 24 points, achieving a top score of 1470. Similarly, on WebDevArena, evaluating web development skills, it rose 35 points to lead with a score of 1443. The model also excels on the Aider Polyglot benchmark, testing complex coding capabilities across multiple languages.
The Gemini 2.5 Pro also ranks among the best on benchmarks like GPQA and Humanity’s Last Exam, designed to test advanced reasoning, math, science, and general knowlege. Google is emphasizing technical excellence with this release,notably in code generation and scientific reasoning,as it competes in the rapidly evolving AI landscape.
Developers can now begin building with the upgraded Gemini 2.5 Pro through Google AI Studio and Vertex AI. The enhanced AI model is also being rolled out in the Gemini app. Google anticipates full general availability of Gemini 2.5 Pro in the coming weeks. The company stated that changes were made to the model based on developer feedback.
What’s next
Google’s continued investment in its Gemini family of AI models underscores its commitment to leading in AI innovation.the company first unveiled Gemini 2.5 in March as a model capable of reasoning, moving beyond simple classification and prediction.