Image AI Models: ChatGPT and Gemini Lead Downloads

Image AI Models Drive the App Market

The launch of image-focused artificial intelligence models is transforming the mobile app market. According to a recent report from app intelligence provider Appfigures, these models have generated 6.5 times more downloads than updates to traditional models. This shift marks a break from early trends, where innovations in conversational models and voice interfaces dominated demand.

For instance, the ChatGPT and Gemini apps each added tens of millions of new downloads following the introduction of their respective image models. The Nano Banana model from Gemini, in particular, led to over 22 million additional downloads in the 28 days following the launch of the Gemini 2.5 Flash image model last August. This launch quadrupled the app's downloads during that period, according to data provided by Appfigures.

Remarkable Performance of ChatGPT

ChatGPT has also seen notable success with its image model GPT-4o, adding over 12 million additional installations in the 28 days following its launch in March of last year. This figure represents about 4.5 times more downloads than its previous models GPT-4o, GPT-4.5, and GPT-5, as highlighted by Appfigures.

Other model launches have followed similar trends, albeit on a more modest scale. For example, Meta AI's introduction of its AI video feed, Vibes, added approximately 2.6 million additional downloads in the 28 days following its launch in September 2025. Although technically a video model, Vibes ultimately concerns visual content, not just text.

Impact on App Revenues

However, the Appfigures report warns that the increase in downloads does not always translate into a rise in mobile revenue. New launches of image models give users a reason to install the app and try its enhanced image generation capabilities, but this does not necessarily mean they will convert into paying subscribers.

For example, while the Nano Banana model produced a significant increase in downloads, it generated only $181,000 in estimated consumer spending during the 28 days following its launch. In contrast, ChatGPT successfully turned the increased attention into real revenue. OpenAI's image generation model 4o led to approximately $70 million in estimated consumer spending in the 28 days following its launch, compared to its previous base.

The Exception of DeepSeek

Appfigures' analysis also examined DeepSeek, although this case does not fit the typical pattern of image models. After the launch of DeepSeek R1 in January 2025, the app recorded 28 million downloads. This success is attributed to the curiosity sparked by its low-cost AI model training techniques, rather than a specific image model. This case underscores how curiosity can drive downloads, although in this instance, the interest is not related to an image model.