Gemini 2.0 Flash and Flash Thinking: the new SOTA models for the agentic era

This podcast interviews Logan Kilpatrick, Google AI Studio lead, about Google's Gemini 2.0 models, focusing on pricing strategies (Flash, Flashlight, Pro), the "experimental" model release process, and the future of reasoning models. The discussion covers the technical aspects of model development, including the shift from scaling parameter size to inference time compute, and the integration of reasoning capabilities into the Gemini models. Kilpatrick highlights the importance of real-time multimodal AI experiences, exemplified by Google AI Studio's live demo, and discusses the potential for integrating AI into existing communication platforms like text and email. A key takeaway is that Google prioritizes providing developers with high-performance models at competitive prices, while also pushing the boundaries of AI capabilities through ongoing research and development. The interview also touches upon Google's collaboration with other companies, such as Cursor, and the future direction of AI model development.