Google has unveiled Gemini 2.5 Pro Experimental, its most advanced artificial intelligence (AI) model to date, designed to tackle increasingly complex problems with enhanced reasoning and coding capabilities.
- Enhanced Reasoning: The model leads in advanced reasoning benchmarks, including math and science assessments like GPQA and AIME 2025. It also achieved a state-of-the-art score of 18.8% on Humanity’s Last Exam, a dataset designed to evaluate human-level knowledge and reasoning.
- Advanced Coding Capabilities: Gemini 2.5 Pro excels in creating visually compelling web applications and agentic code applications, along with code transformation and editing. On SWE-Bench Verified, an industry standard for agentic code evaluations, it scored 63.8% with a custom agent setup.
- Multimodal Understanding: The model can process and understand inputs across various formats, including text, audio, images, and video, enhancing its versatility in handling diverse data types.
- Extended Context Window: Initially offering a 1-million token context window, Gemini 2.5 Pro is set to expand to 2 million tokens, allowing it to handle extensive datasets and entire code repositories effectively.
According to Koray Kavukcuoglu, Chief Technology Officer of Google DeepMind, Gemini 2.5 Pro Experimental leads common benchmarks by meaningful margins and showcases strong reasoning and code capabilities. The model tops the LMArena leaderboard, which measures human preferences, indicating a highly capable model equipped with high-quality style. It also shows strong reasoning and code capabilities, leading on common coding, math, and science benchmarks.
The introduction of Gemini 2.5 Pro Experimental is part of Google’s ongoing efforts to enhance its AI offerings and maintain a competitive edge in the rapidly evolving AI landscape. The model is currently available in Google AI Studio and for Gemini Advanced subscribers, with plans for integration into Google’s cloud-computing product, Vertex AI, in the near future.
This development comes amid intensifying competition in the AI sector, with tech giants like Microsoft and Google-parent Alphabet vying to integrate more advanced reasoning capabilities into their AI models. Microsoft has announced the launch of two new reasoning agents, Researcher and Analyst, for their Microsoft 365 Copilot software, enhancing its ability to analyze vast amounts of data efficiently. These agents will be accessible to customers starting in April, as part of a program granting early access to new features.
As AI technology continues to advance, models like Gemini 2.5 Pro Experimental are poised to play a pivotal role in shaping the future of AI applications, offering users more sophisticated and personalized experiences across various platforms.
