Google has introduced an updated version of its AI model, Gemini 2.5, which boasts improved performance in various areas, including reasoning, multimodality, and context-aware capabilities. The new model combines a significantly enhanced base model with advanced post-training to provide better overall results.
Gemini 2.5 Pro is the latest release, which outperforms its competitors on common AI benchmarks that measure understanding, mathematics, coding, and other capabilities. This includes models from OpenAI, Anthropic, xAI, and DeepSeek. The new model is available for Gemini Advanced subscribers in the app’s model dropdown menu or accessible through Google AI Studio.
Google highlights the improved reasoning capabilities of Gemini 2.5 Pro, which enable tasks to be processed step-by-step, leading to more informed decisions and better answers for complex prompts. This allows the model to handle more complex problems and support even more capable agents.
A demo video showcases Gemini 2.5 Pro using its reasoning capabilities to program a video game based on a single prompt. Google’s DeepMind CEO Demis Hassabis calls it “an awesome state-of-the-art model” with significant improvements in multimodal reasoning, coding, and STEM. The company is also upgrading its context window size to 2 million tokens, expected to further enhance data processing capabilities.
Source: https://www.theverge.com/news/635502/google-gemini-2-5-reasoning-ai-model