Google releases Gemini 3 Flash, promising improved intelligence and efficiency (arstechnica.com)

Google has officially launched Gemini 3 Flash, an enhanced version of its generative AI model that promises superior intelligence and efficiency compared to previous iterations. This new model integrates into the Gemini app and search, and is immediately accessible to developers via the Gemini API, Vertex AI, AI Studio, and Antigravity. Notably, Gemini 3 Flash has outperformed its predecessor, the 2.5 Flash, in key academic benchmarks, showcasing substantial improvements in areas like domain-specific knowledge and coding skills. For instance, it scored 33.7% on Humanity’s Last Exam, a significant leap over the prior model, and stands just behind the Gemini 3 Pro in several tests, indicating its robust capabilities. Significantly, Gemini 3 Flash not only enhances performance but does so with improved efficiency, processing workloads three times faster than the older version. The cost for developers has been adjusted; accessing one million input tokens will now cost $0.50, compared to $0.30 with the 2.5 Flash. While the pricing has increased, the overall improvements in accuracy, speed, and coding proficiency position Gemini 3 Flash as a compelling option in the AI landscape, maintaining the key features that allow for interactive simulations and multimodal content generation, thus broadening its appeal within the AI/ML community.