Gemma 3 is a newly released multimodal AI model that integrates both image and text processing capabilities, supporting an impressive 128K token context across over 140 languages. It is available in multiple configurations ranging from 1 billion to 27 billion parameters, with open weights for commercial use. The announcement also highlights its optimization for AMD ROCm, making it accessible for users who depend on this technology. Comments from users reflect a mix of enthusiasm for the technical advancements, such as improvements in inference efficiency and the innovative approach to memory usage, while also voicing concerns about model limitations and the implications of content moderation policies. Users have pointed out that despite its enhancements, there are still areas where the model struggles, such as common sense reasoning and generating content without biases.