Following the launch of Gemini 3 Pro and Gemini 3 Deep Think, Google, the tech leader based in Mountain View, has unveiled Gemini 3 Flash, its latest lightweight AI model. This model will serve as the default in the Gemini app and AI mode for searches, taking over from the 2.5 Flash version. It is anticipated to surpass the 2.5 Flash model as well as Gemini 2.5 Pro in both performance and reasoning capabilities.
Gemini 3 Flash: A Quicker, Smarter, and More Efficient AI
According to a blog post from Google, the Gemini 3 Flash achieved a score of 33.7% in Humanity’s Last Exam, compared to the 11% scored by the Gemini 2.5 Flash. The new AI model also recorded 81.2% in the MMMU-Pro test, which evaluates AI understanding and reasoning through various inputs including text and images.
With this innovative AI model, Google has maintained the core intelligence and reasoning capabilities from Gemini 3.0 while enhancing its speed and efficiency characteristic of lightweight AI models. The multimodal reasoning features of Gemini 3 Flash allow it to effectively analyse visuals and sounds, enabling users to obtain answers more swiftly. Google provided an illustration of its functionality, explaining that users can prompt Gemini to interpret their videos and images, converting that content into an actionable plan in mere seconds.
Moreover, users can verbally express their ideas in simple language, and the model can convert these thoughts into a functioning application in a few minutes. The rollout of the new Gemini model has commenced for all users in both the Gemini app and AI Mode for Search. It is also accessible for preview in the Gemini API within Google AI Studio, Google Antigravity, Vertex AI, and Gemini Enterprise.
