Google has launched a new artificial intelligence image model called Gemini 2.5 Flash Image, which it says is its best image creation and editing tool so far.
The model was secretly tested earlier under the funny name Nano Banana, and it had already reached the top position on the community ranking website LMArena before today’s official release.
One of the main problems people had with older AI models was that characters in images would change or get distorted when making edits. For example, if you tried to change someone’s t‑shirt color, their face or body could look very different from the original.
Google claims the new Gemini 2.5 fixes this problem. It can now keep the person looking the same while allowing changes like adding a hat or changing clothing colors, without losing the original look.
The new model is also much faster and allows very precise editing of certain parts of an image. Users can also edit in steps, meaning you can ask for one change, then another, and the AI will remember the character details with every edit. Another fun feature is image blending, where users can mix two different images into one. The results can vary depending on what images are used, but it can create unique outputs.
The model can be tried out by anyone through the Gemini app. At the same time, developers and businesses can use it through the Gemini API, Google AI Studio, and Vertex AI. The cost has also been revealed. The price is $30 per million tokens, and creating one image takes about 1,290 tokens. This means one image costs around $0.039, or about ₹3.5 in India.
In a blog post, Google confirmed that Nano Banana on the LMArena platform was actually Gemini 2.5 Flash Image being tested quietly before its launch. Other companies have also done similar things in the past. For example, OpenAI tested its GPT‑4.1 model secretly in the same way before announcing it officially.
This launch is important because the competition in AI image generation is currently very strong. Tools like DALL·E 3 from OpenAI, MidJourney, and Stable Diffusion XL are already popular with designers, artists, and everyday users. Google now hopes that Gemini 2.5, with its better speed and strong character consistency, will stand out in the market.
Many experts believe the improvements in accuracy will be very useful for areas like advertising, media, design, and content creation, where keeping an image close to the original is very important.
The timing of this release is also notable. Reports suggest that even Apple may be thinking of using Gemini AI to give a smarter upgrade to Siri in upcoming devices, including its first foldable iPhone. At the same time, Google is also growing its Gemini platform with more ways for developers and companies to use it, especially through enterprise tools like Vertex AI.
With Gemini 2.5 Flash Image, Google is showing that it wants to be a major player in the world of AI image generation. By solving the problem of character distortion and giving faster and more flexible editing options, the company is hoping to attract both everyday users and professionals.
While it is still early to say how it will compare with MidJourney or DALL·E 3 in real use, the first response from users is already positive, and many believe this could be Google’s biggest step forward in creative AI so far.