According to Sundar Pichai, the Gemini 2.5 Computer Use model is now available in the Gemini API with lower latency and "sets a new standard on multiple benchmarks" (source: Sundar Pichai on X, Oct 7, 2025). According to Sundar Pichai, the model can interact with the web by scrolling, filling forms, and navigating dropdowns, described as an important next step toward general-purpose agents (source: Sundar Pichai on X, Oct 7, 2025). According to Sundar Pichai, developers can access these capabilities via Google AI Studio and Vertex AI through the API (source: Sundar Pichai on X, Oct 7, 2025).

According to DeepLearning.AI, Google has introduced Gemini 2.5 Pro Experimental, marking the debut of its new Gemini 2.5 family. This advanced model, designed with enhanced reasoning and coding capabilities, is trained using reinforcement learning to generate hidden reasoning steps. It currently tops the Chatbot Arena leaderboard, demonstrating a significant leap in AI performance and potential applications in cryptocurrency trading automation. The model's ability to process complex reasoning tasks could lead to more precise trading algorithms and decision-making systems.