Google DeepMind is reportedly developing a new AI model called Gemini Ultra, following the release of Gemini Pro 2.5. The Gemini Pro 2.5 model has been noted for its strong performance, especially when compared to OpenAI's O3 and O4-mini models, raising expectations for the upcoming Gemini Ultra. Gemini 2.5 Flash, a recent iteration, features hybrid reasoning capabilities, a 1 million token context window, multimodal input support, and native search functionality, all optimized for speed and cost-efficiency. However, recent reports indicate that one of Google's latest Gemini AI models has scored worse on safety metrics compared to its predecessor, drawing attention from technology and cybersecurity news outlets such as TechCrunch. This development highlights ongoing challenges in balancing AI advancement with safety considerations.