Image AI models now drive app growth, beating chatbot upgrades
Source: TechCrunch
Image model releases are driving growth for AI mobile apps, generating 6.5× more downloads than traditional model updates, according to a new report from app‑intelligence provider Appfigures.
Shift from conversational‑only upgrades
Earlier, the release of new models powering conversational experiences—such as the ones highlighted in these TechCrunch articles—drove demand, especially when paired with features like a voice‑chat interface.
- OpenAI makes ChatGPT more direct, less verbose
- OpenAI’s newest model is GPT‑4o
- ChatGPT’s new face is a black hole (voice chat)
Major download spikes from image model launches
Google Gemini
The release of the Nano Banana image model added 22 + million downloads in the 28 days after the launch of the Gemini 2.5 Flash image model (August). This lifted the app’s downloads by more than 4× over that period.
Image Credits: Appfigures
OpenAI ChatGPT
The introduction of the GPT‑4o image model in March generated 12 million incremental installs in the following 28 days, roughly 4.5× more downloads than the GPT‑4o, GPT‑4.5, and GPT‑5 releases combined.
Image Credits: Appfigures
Other visual‑model releases
Meta AI – Vibes
The AI video feed Vibes added an estimated 2.6 million incremental downloads in the 28 days after its September 2025 release. Although it is a video model, it falls under the broader “visual content” category.
Image Credits: Appfigures
Revenue impact
The report notes that higher download numbers do not always translate into increased mobile revenue.
- Nano Banana drove only $181,000 in estimated gross consumer spending during its 28‑day window, despite the large download spike.
- Meta AI’s Vibes generated additional downloads but no meaningful revenue.
- ChatGPT’s GPT‑4o image model produced an estimated $70 million in gross consumer spending over the same period, turning the download surge into substantial dollars.
Outlier case: DeepSeek
DeepSeek’s R1 model sparked 28 million downloads after its January 2025 release, but the surge was driven more by curiosity about the company’s low‑cost training techniques than by an image‑model upgrade. This highlights that not all download spikes are tied to visual‑model releases.