Investing.com -- Alphabet Inc's Google (NASDAQ:GOOGL) has revealed Gemini 2.5, representing a significant leap forward in its artificial intelligence capabilities. The initial release, Gemini 2.5 Pro Experimental, showcases remarkable advancements in reasoning and coding, substantially outperforming competitors across major industry benchmarks.
At the core of Gemini 2.5's innovation is its advanced reasoning system, allowing the AI to effectively "think through" problems before generating responses. This deliberate reasoning process—analyzing information, drawing logical conclusions, and incorporating contextual nuances—results in significantly higher accuracy and performance.
The model builds upon Google's earlier work with reinforcement learning and chain-of-thought prompting, which first materialized in Gemini 2.0 Flash Thinking. The 2.5 series combines an enhanced foundation model with sophisticated post-training techniques, enabling it to tackle more complex challenges and support more capable, context-aware applications.
Gemini 2.5 Pro Experimental has claimed the top position on the LMArena leaderboard, which measures human preferences for AI interactions. The model demonstrates exceptional reasoning abilities, particularly excelling in mathematics and scientific reasoning benchmarks like GPQA and AIME 2025—without relying on costly test-time enhancement techniques.
Notably, the model achieved a state-of-the-art score of 18.8% on Humanity's Last Exam, a challenging dataset designed to test the boundaries of knowledge and reasoning capabilities.
Google has placed special emphasis on enhancing Gemini's coding abilities. The 2.5 Pro model excels at creating visually sophisticated web applications and complex code structures, while also demonstrating proficiency in code transformation and editing tasks. On SWE-Bench Verified, the industry's standard for evaluating code-related AI capabilities, Gemini 2.5 Pro achieved an impressive 63.8% score with a customized agent configuration.
Building on previous generations, Gemini 2.5 features native multimodal processing and an expansive context window of 1 million tokens (with plans to increase to 2 million). This allows the model to process and comprehend massive datasets from diverse sources, including text, audio, images, video, and entire code repositories with improved performance over earlier versions.
Gemini 2.5 Pro is currently available through Google AI Studio and in the Gemini app for Gemini Advanced subscribers. Google plans to expand access via Vertex (NASDAQ:VRTX) AI soon, with pricing details to be announced in the coming weeks. The new pricing structure will enable scaled production use with higher rate limits.
This content was originally published on Investing.com
