Multiverse Computing has successfully created a smaller, de-censored version of DeepSeek R1, a powerful AI model originally developed in China. This new model, DeepSeek R1 Slim, retains nearly the same performance while being 55% smaller, allowing for responses free from the constraints of Chinese censorship. It answers sensitive questions similar to Western AI models, thus highlighting significant advancements in AI technology.
Additionally, Google has introduced Gemini 3, which is a substantial overhaul of its multimodal capabilities. This upgrade enhances reasoning skills and enables functionalities across various media types, including text, voice, and images. A standout feature of Gemini 3 is the Gemini Agent, designed for managing complex tasks within the app by integrating with Google services like Calendar and Gmail. This model signifies a leap toward more interactive AI experiences, dealing efficiently with everyday tasks.
👉 Pročitaj original: MIT Technology Review
