The Gemini family expands with 2.5 reasoning models. 2.5 Pro Experimental is, according to Google, its most intelligent model ever.
Google has launched its most advanced AI model to date: Gemini 2.5 Pro Experimental, a reasoning model that thinks before providing an answer.
Reasoning as the Future of AI
Increasingly, companies like OpenAI are adding reasoning capabilities to their AI. They take more time and use more computing power to provide an answer that leads to better results. Google now confirms that all new AI models will have this feature as standard.
Google says that Gemini 2.5 Pro excels in building visual web apps and supporting agentive coding scenarios. It outperforms models from OpenAI and DeepSeek on some benchmarks. In a test on agentive AI, it scores 63.8 percent, just below Anthropic’s Claude 3.7 Sonnet (70.3%).
Large Memory, Great Potential
The model can process one million tokens or about 750,000 words at once, with an upgrade to two million tokens on the horizon. This outperforms most comparable models.
Gemini 2.5 Pro Experimental is now available in Google AI Studio and via the Gemini app for Gemini Advanced subscribers (21.99 euros per month). Google will soon make Gemini 2.5 Pro available in Vertex AI. Prices for API usage are not yet known.