Google just rolled out Gemini 2.5 Flash Image, calling it their most advanced AI for creating and editing pictures—think crisp mountain skylines or detailed city streets. The model generates results more quickly, lets you fine-tune individual elements, and keeps characters looking the same—right down to details like a familiar scar or hairstyle. It first appeared on LMArena, a crowdsourced AI ranking site, under the playful alias Nano Banana, where it shot to the top and drew praise for crisp, polished edits.
Key Features and Improvements:
The new model makes editing smoother, keeping a character’s look consistent even when you swap a blue shirt for a red one or add a hat—without strange distortions. It also includes an image-blending tool, letting you merge two pictures into a single frame—like overlaying a sunset on a city skyline—though how well it works depends on your inputs and prompts. It also allows multi-turn editing, so you can ask for several changes in a row while keeping the original character intact—like adjusting a line’s tone, then tweaking a single word without losing its voice. Google says the model runs fast and produces top-tier results, such as crisp, clear images in seconds.
Gemini 2.5 Flash Image: Access and Pricing
You can try Gemini 2.5 Flash Image in the Gemini app, while developers can access it through the Gemini API, Google AI Studio, or—at the enterprise level—Vertex AI. It costs $30 for every million output tokens, and since each image uses 1,290 of them, that works out to about $0.039—or roughly ₹3.5—for a single image. Google confirmed that Nano Banana on LMArena was actually the Gemini 2.5 Flash Image—much like how OpenAI quietly tested GPT-4.1 behind the scenes.
