Apple has launched its Foundation Models framework, enabling developers to access its core AI models locally on devices without requiring cloud connectivity or incurring API costs. This framework supports text and image processing in 15 languages and is optimized for Apple Silicon, emphasizing privacy and offline capabilities. Meanwhile, Google is advancing its AI offerings with the upcoming release of Gemini 2.5 models, including the Gemini 2.5 Flash Lite Preview scheduled for June 17 and the Gemini 2.5 Pro set for June 19. The Gemini 2.5 Pro model has already demonstrated strong performance, topping public leaderboards with achievements such as the best model at 192,000 tokens on Live Fiction, number one on SimpleBench with a 62.4% score, and leading in document processing tasks. Additionally, a new Google DeepMind model named "Prowlridge" has appeared in AI testing arenas like WebArena and LM Arena, signaling ongoing innovation. Google’s AI suite, including Gemini 2.5 Pro, Veo 3, and Imagen 4, positions the company as a leading vertically integrated AI player, ranking equal number two in intelligence on the Artificial Analysis Intelligence Index. This competition highlights a dynamic AI landscape with Apple focusing on privacy-centric, on-device AI access and Google pushing the boundaries of model performance and integration.