Google on Tuesday rolled out Gemini 2.5 Flash Image, a state-of-the-art image-generation and editing model developed by Google DeepMind. The system—known during public testing as “Nano Banana”—is now live for all users of the Gemini mobile app and is available to developers through the Gemini API, Google AI Studio and Vertex AI. Gemini 2.5 Flash Image blends multiple photos, executes precise natural-language edits and, according to Google, preserves character likeness across sequential changes. The company says the tool embeds an invisible SynthID watermark and metadata to flag AI-generated content. Before the disclosure, the model anonymously topped the crowdsourced LMArena leaderboard, finishing roughly 171 Elo points ahead of its nearest rival and attracting more than 2.5 million direct votes out of some 5 million cast in the last two weeks. It also ranked first in the site’s text-to-image category. The service is free inside the consumer Gemini app, while enterprise use is priced at $30 for 1 million output tokens—about four cents per image. Google is distributing the model through partners such as OpenRouter and fal.ai, and Adobe said it is integrating the technology into Firefly and Express, granting premium users unlimited generations until 1 September. The release escalates Google’s competition with OpenAI’s GPT-4o and Microsoft-backed Midjourney in visual generative AI. Shares of Adobe, whose Photoshop and Firefly businesses overlap with the new offering, slipped 1.8% in afternoon trading.