Google on Thursday launched the latest version of its Nano Banana image generation model, which promises to improve upon last November’s Nano Banana Pro and the older Nano Banana 1 in almost every aspect, especially speed.Nano Banana’s official name is Gemini 3.1 Flash Image, and the ‘Flash’ in there gives away its pedigree. Google’s Flash models have always focused on speed over raw intelligence, but for Nano Banana 2, the company is also bringing many features from Nano Banana Pro to this faster model.This means, for example, that the new model will also pull from Gemini’s real-world knowledge base and can access real-time data and images from the web.“This deep understanding also helps you create infographics, turn notes into diagrams, and generate data visualizations,” Google explains in today’s announcement.It is also far better at rendering text, which was where most of the first-generation image generation models struggled. Today’s models are already quite good at this, and Nano Banana 2 takes it a step further by translating and localizing text on command.Image credit: Google.The older Pro model isn’t going away, at least for Google AI Pro and Ultra subscribers who can still use it for specialized tasks.As Google notes in its announcement, Nano Banana 2 is meant to “close the gap between speed and visual fidelity.”Image credit: Google.Among the improvements from Nano Banana 1 are better subject consistency, for example. The model can now track up to five characters and 14 objects across a single workflow, which should make it easier to storyboard ideas, among other things. This matches what Nano Banana Pro could do at launch.Lighting, textures, and details should also now be more realistic.Like Nano Banana Pro, it can take multiple images as input and combine them into a single coherent result.Image credit: Google.Google says the new model is also better at following instructions, especially when it comes to complex requests, and makes it easier to generate multiple versions of an image at different aspect ratios and resolutions.Image credit: Google.The model went live in the Gemini app shortly before the official announcement and after testing it a bit, the new model is definitely much faster. It still occasionally makes some interesting mistakes (check out the travelers on top of the jet bridge in the image below, the changing color of the logo below, and the mangled United wordmark), but it is also very good at fixing those when prompted.Image generated by Nano Banana 2 in the Gemini app. Prompt: “Create an image of the New Stack logo on a 737 at an airport gate.”Nano Banana 2 will become the default in the Gemini app, as well as in Google Search’s AI mode and for Lens. For developers using Google’s API, it’s now available in preview in AI Studio, the Gemini API, and the company’s Antigravity next-gen IDE. It’ll also be the default for Flow, Google’s video generation tool, and Google Ads. The post Google’s Nano Banana 2 promises Flash speeds with Pro results appeared first on The New Stack.