Google has unveiled two new models in the Gemini family: Gemini 3.6 Flash and Gemini 3.5 Flash-Lite, which both achieve significant improvements in task efficiency. Specifically, Gemini 3.6 Flash reduces the average time per task by over 50% to 1.3 minutes while maintaining its Intelligence Index score of 50, matching its predecessor, Gemini 3.5 Flash. In contrast, the newly released Gemini 3.5 Flash-Lite enhances its intelligence score by 11 points to 36, almost cutting task time in half to 0.6 minutes. These updates align with a common trend among leading AI labs to refine model speed and efficiency for practical applications, highlighting ongoing developments in AI benchmarking practices that evaluate these enhancements.

Google: Google is a leading technology company focused on internet services, cloud computing, and artificial intelligence development. Its AI efforts center on building advanced multimodal models that integrate text, image, video, and speech capabilities. The company has released the latest Gemini model updates through its research division to enhance task efficiency.
@GoogleDeepMind: @GoogleDeepMind serves as the primary online channel for Google DeepMind, the company’s dedicated AI research laboratory. It shares announcements, benchmark results, and technical details on new model releases. The account highlighted the launch of Gemini 3.6 Flash and Gemini 3.5 Flash-Lite with supporting performance analysis.
Gemini 3.6 Flash: Gemini 3.6 Flash is a high-reasoning model in Google’s Gemini family optimized for advanced problem-solving and agentic tasks. It preserves intelligence benchmarks from prior versions while prioritizing faster output and token efficiency. This model forms part of the recent Gemini family updates announced for improved performance in real-world applications.
Gemini 3.5 Flash-Lite: Gemini 3.5 Flash-Lite is an efficient variant in the Gemini series designed for lighter workloads with strong agentic capabilities. It delivers meaningful intelligence gains over earlier lite models alongside substantial reductions in task processing time. The model was included in the latest release to expand options for cost-effective AI deployment.

AI Model Iteration: Leading AI labs regularly release updated model variants that refine speed and efficiency for practical use cases.
Benchmarking Practices: Independent evaluations track AI progress across intelligence indices and task completion metrics to compare new releases.