The new Gemini 3.5 Flash-Lite model achieves 350 output tokens per second and costs $0.09 per task on Artificial Analysis, making it significantly more efficient. It is currently rolling out in Google Search and outperforms Gemini 3 Flash in coding and computer use, indicating improved agentic capabilities.
Google's aggressive pricing and performance for Gemini 3.5 Flash-Lite will drive down inference costs, increasing demand for underlying compute infrastructure.