Google releases the image generation and editing model Nano Banana 2.1
BIBIBI
AT A GLANCE
PANews states that Google released Nano Banana 2.1, and that it is now available in the Gemini app and on API.
Article
Google launches image model Nano Banana 2.1, with improved text rendering capabilities
PANews October 6 reports that, according to the model card on the official Google DeepMind website, Google has released the image generation and editing model Nano Banana 2.1, built on Gemini 3.6 Flash, with significant improvements over its predecessor in image quality, prompt adherence, character consistency, and text rendering, supporting outputs of up to 4K at most. In preference evaluations, its thinking mode scored above 1050, higher than Nano Banana Pro's 935. The model is now available in the Gemini app and on API.
PANews, citing the model card on the official Google DeepMind website, reports that Google has released the image generation and editing model Nano Banana 2.1。
02
The original text states that Nano Banana 2.1 is built on Gemini 3.6 Flash.
03
The original text states that the model has significantly improved over its predecessor in image quality, prompt adherence, character consistency, and text rendering.
04
The original text states that the model supports outputs of up to 4K.
05
The original text states that, in preference evaluations, the model's thinking mode scored above 1050, while Nano Banana Pro scored 935。
06
The original text states that the model is now available in the Gemini app and on API.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
This is a report about the release of an image model. According to the original text, Nano Banana 2.1 can both generate and edit images, with key improvements in creating images based on prompts, maintaining consistent character appearances, and generating text within images, while supporting outputs of up to 4K.
Why it matters to readers
If the capabilities and availability described in the original text are accurate, users can use the model in the Gemini app, and developers can integrate it through API; improvements in text rendering and character consistency may broaden its applicable use cases.
Beginners may want to focus on whether it is easier to generate images according to text requirements and whether text within images is clearer and more accurate.API It is primarily intended for developers who want to integrate the model into their own products.
Risks and unknowns
The evidence consists only of PANews' account of the Google DeepMind model card; it does not provide the model card's original link, so the release and availability information cannot be independently verified from this evidence.
The testing methodology, sample scope, and comparison conditions for the claims of "significant improvement" and the preference evaluation scores are not explained in the original text.
Gemini Application and API its available regions, pricing, quotas, and 4K output conditions are not specified.
No independent evaluations or actual user feedback were provided in the source text.
Related Developments
Loading event timeline…
Related concepts
Prompt adherence
This term is not in the glossary yet. Browse related concepts in the glossary.
