Recent data from app intelligence firm Appfigures reveals that image model updates are now the primary growth engine for AI mobile applications, generating 6.5 times more downloads than traditional model releases. This represents a significant shift from earlier periods, when conversational model updates and features like voice chat interfaces were the main drivers of user demand.

For example, both ChatGPT and Gemini experienced substantial download spikes after launching their respective image models. Google's Gemini saw over 22 million additional downloads in the 28 days following the introduction of its Nano Banana image model, which came alongside the Gemini 2.5 Flash release last August. This launch boosted the app's download volume by more than four times during that period.

Similarly, ChatGPT gained more than 12 million incremental installs in the month after its GPT-4o image model debuted in March of last year. According to Appfigures, this represented roughly 4.5 times more downloads than what the app saw from its GPT-4o, GPT-4.5, and GPT-5 model releases combined. Other AI apps followed comparable trends, though on a smaller scale. Meta AI's introduction of its AI video feed feature, Vibes, added an estimated 2.6 million incremental downloads in the 28 days following its September 2025 release. (While technically a video model, it still centers on visual content rather than text.)

However, the report cautions that increased download numbers don't automatically translate to higher mobile revenue. Instead, new image model releases give users a reason to install the app and explore its upgraded image generation capabilities—but that doesn't guarantee they will become paying subscribers. For instance, Appfigures noted that Nano Banana generated only about $181,000 in estimated gross consumer spending during the 28-day window after its release, despite producing a larger download spike than ChatGPT's 4o image model launch. Meta AI's Vibes rollout also led to more downloads but produced no meaningful revenue. Among the three, only ChatGPT successfully converted the heightened attention into actual dollars. OpenAI's 4o image generation model brought in an estimated $70 million in gross consumer spending over the 28 days following its launch, compared to its previous baseline, according to Appfigures.

The analysis also examined DeepSeek, but it didn't follow the same pattern. Although DeepSeek R1 drove 28 million downloads after its January 2025 release, this wasn't a typical model comparison event. Instead, it marked DeepSeek's breakout moment, when the app went from relative obscurity to an overnight sensation as the tech industry discovered the innovative techniques it used to train its AI models at a fraction of the cost of competitors. This case illustrates how curiosity can fuel downloads—though in this instance, the interest wasn't tied to an image model.