Gemini 3.1 Flash (Nano Banana 2) Spotted Live in Gemini Ahead of Official Release

Google's next-generation lightweight AI model, Gemini 3.1 Flash, has been spotted live in the Gemini user interface ahead of any official release. Internally referred to by the quirky codename 'Nano Banana 2', the model is already loaded and appears as a selectable option for users, strongly suggesting a staged or accidental early rollout. This discovery, reported by a user on social platform Reddit, indicates that Google's deployment process is in its final stages. The appearance of Gemini 3.1 Flash follows the successful launch of its predecessor, Gemini 1.5 Flash, which was designed for high-speed, low-cost inference on tasks like summarization, chat applications, and data extraction where the massive context window of larger models isn't required.

While technical specifications for Gemini 3.1 Flash remain unconfirmed, its predecessor set a high bar with a 1 million token context window and significantly lower latency and cost compared to the Pro model. The early appearance suggests Google is preparing to announce performance improvements, potentially in speed, accuracy, or efficiency for its 'Flash' line. For developers and businesses, this signals that a new, more capable tool for building scalable AI applications is just around the corner. The imminent release will likely intensify competition in the mid-tier AI model market, challenging offerings like OpenAI's GPT-4 Turbo and Anthropic's Claude 3 Haiku.