Google Gemini’s AI image model has received a significant upgrade with the launch of Gemini 2.5 Flash Image, a new state-of-the-art image generation model designed to deliver faster, more accurate, and highly detailed visual outputs. The update enhances the model’s ability to understand complex prompts and produce realistic, contextually appropriate images, marking a major leap in performance and efficiency. With improved speed and precision, the new model is optimized for real-time applications, creative workflows, and interactive experiences. Google is positioning Gemini 2.5 Flash Image as a powerful tool for developers, designers, and content creators seeking advanced image generation capabilities. The model’s name, “bananas,” reflects its enhanced performance and versatility, signaling a bold step forward in the evolution of generative AI for visual content.

Google has launched Gemini 2.5 Flash Image, a new AI-powered image generation and editing model, rolling out to all users in the Gemini app and developers via the Gemini API, Google AI Studio, and Vertex AI. The update, internally known as “nano-banana” during testing, is designed to give users unprecedented control over photo edits with natural language commands, aiming to close the gap with OpenAI’s image tools and boost user engagement. The model excels at precise, localized edits while preserving facial and object consistency—something many competitors struggle with. For example, changing a person’s shirt color or removing an unwanted object from a photo now maintains realistic details and avoids distortions. It also supports multi-image fusion, allowing users to combine multiple photos into a single cohesive image—such as placing a product into a new environment or merging a color palette with a room photo. The model leverages Gemini’s deep world knowledge, enabling it to interpret complex prompts, understand hand-drawn diagrams, and generate contextually accurate visuals. Google emphasizes that the model was built with real-world consumer use cases in mind, such as visualizing home renovations or designing product mockups. It also supports multi-turn conversations, letting users refine edits iteratively. Developers can now build custom AI apps in Google AI Studio with pre-built templates, enabling rapid prototyping and deployment. Priced at $30 per million output tokens—about $0.039 per image—Gemini 2.5 Flash Image is available in preview, with plans to stabilize in the coming weeks. Google has partnered with OpenRouter.ai and fal.ai to extend access to millions of developers globally. Despite past missteps with AI-generated content—like historically inaccurate images that led to a temporary rollback—Google says it has improved safeguards. The model now blocks non-consensual intimate imagery and applies invisible SynthID watermarks to all AI-generated or edited images, helping combat deepfakes. However, users must still be cautious, as these identifiers may not be visible to casual viewers. The update comes amid fierce competition in the generative AI space. OpenAI’s GPT-4o image generator fueled a surge in ChatGPT usage, reaching over 700 million weekly users. Google’s Gemini, with 450 million monthly users, is aiming to grow its user base with stronger creative tools. The new image model could be a key differentiator, especially as Meta recently announced plans to license Midjourney’s AI models and Black Forest Labs continues to lead in benchmark performance. Google’s product leads, including Nicole Brichtova from Google DeepMind, stress that the model is not just about speed and cost but about quality, precision, and creative control. The company is actively improving long-form text rendering, character consistency, and factual accuracy based on user feedback. With its enhanced editing capabilities, world knowledge, and developer accessibility, Gemini 2.5 Flash Image marks a significant step forward in AI-powered visual creation. While challenges around content safety and authenticity remain, the update positions Google to better compete in the rapidly evolving generative AI landscape.