Gemini 2.5 Pro, announced in March as the most intelligent model to date, has now been further enhanced with updates that solidify its leadership in various domains. Beyond excelling in academic benchmarks, Gemini 2.5 Pro leads in web development and learning applications. New features include native audio output for improved conversational interactions, advanced security measures, and capabilities from Project Mariner for computer-related tasks. Additionally, an experimental mode called Deep Think will be introduced to boost reasoning for complex math and coding challenges. For developers, Google continues to refine the Gemini experience by introducing thought summaries in both the Gemini API and Vertex AI, offering greater transparency. Extended "thinking budgets" give users more control over the model's reasoning depth, while support for MCP tools has been added to the API and SDK. These updates aim to empower developers to create more sophisticated and impactful web applications.

本专栏通过快照技术转载，仅保留核心内容

内容中包含的图片若涉及版权问题，请及时与我们联系删除