This session shares insights and best practices gathered from Google’s collaboration with top Google Cloud enterprises customers to optimize Gemini 3 applications for quality, cost, and scalability. We will deconstruct real-world implementation questions—ranging from complex agentic setups to high-volume data processing—and share the practical lessons learned. The session covers: * Model Behavior: How to adapt your prompting strategy to get the most from Gemini 3’s capabilities * Performance Tuning: Concrete steps to reduce latency and manage token costs without sacrificing response quality. * Resilience and scalability: Designing prompts that are robust against production data distributions. * Advanced Use Cases: Deep dives into implementing advanced use cases, like agents, long-context retrieval, and autoraters. * Migration Tools: methods for updating existing prompts to align with Gemini 3’s instruction-following standards.