Gemini 3 Flash – Everything you need to know (artificialanalysis.ai)

Google has unveiled the Gemini 3 Flash Preview, a new AI model that offers a remarkable combination of cost efficiency and intelligence, being two times cheaper than Gemini 3 Pro Preview with only a slight decrease in the Intelligence Index score. Scoring 71 on the Artificial Analysis Intelligence Index, Gemini 3 Flash represents a significant leap forward from its predecessor, Gemini 2.5 Flash, with a 13-point improvement. This model excels in knowledge and reasoning capabilities, achieving the highest score in the AA-Omniscience benchmark while also performing well in Humanity’s Last Exam and other evaluations. However, it comes at the cost of increased token usage, which surpasses that of other models, including previous versions and competitors, highlighting a trade-off between performance and resource consumption. The technical specifications of Gemini 3 Flash indicate substantial advancements, particularly in its multimodal capabilities, handling text, images, video, and audio inputs effectively. Despite a slowdown in processing speed compared to Gemini 2.5 Flash, it still outperforms many of its peers. The model's ability to support a million-token context window, tool calling, structured outputs, and JSON mode enhances its applicability in various AI tasks. All these features position Google as a continued leader in the AI space, promising innovative applications while retaining a focus on cost efficiency, as demonstrated by its low token pricing structure in comparison to its intelligence level.