Particle.news

Download on the App Store

Google DeepMind Rolls Out Gemini 2.5 Flash Image, Confirms ‘Nano‑Banana’ Origins

Google frames the release around likeness‑preserving edits controlled by natural language with provenance markers baked in.

Image
AI generated image from Gemini 2.5 Flash Image.
Image
Image

Overview

  • The model is available today in the Gemini app for free and paid users, and to developers via the Gemini API, Google AI Studio and Vertex AI.
  • Google identified the top‑rated LMArena editor known as “nano‑banana” as Gemini 2.5 Flash Image, following weeks of anonymous testing.
  • Capabilities include preserving a person’s or pet’s look across edits, multi‑turn workflows, blending multiple photos into one scene and applying one image’s style to objects in another.
  • All outputs include a visible AI label plus Google’s invisible SynthID marker, with usage bounded by policy limits such as bans on non‑consensual intimate imagery.
  • Adobe said it will offer the model in Firefly and Adobe Express, while partners like OpenRouter and fal extend access for developers; pricing works out to about $0.039 per image.