Google DeepMind adds agentic video understanding to Gemini models

Google DeepMind announced a new capability called agentic video understanding, now rolling out across its latest Gemini AI models. The feature is designed to improve how the models interpret and analyze video content.

According to the announcement, the update brings better accuracy alongside lower operating costs and reduced token usage, making video-based AI tasks more efficient. The company positioned this as a step forward in expanding Gemini's multimodal capabilities beyond text and images into more sophisticated video comprehension.