Image model releases are now the dominant growth lever for AI mobile apps, generating 6.5x more downloads than traditional model updates, according to a report from app intelligence firm Appfigures. Google's Gemini added more than 22 million downloads in the 28 days after its Nano Banana image model launched last August, lifting Gemini app installs by more than 4x over that window. ChatGPT added 12 million incremental installs in the 28 days after the GPT-4o image model arrived in March 2025 — roughly 4.5x what its GPT-4o, GPT-4.5, and GPT-5 text releases each pulled in.
The pattern marks a clear shift in what consumers actually open the App Store for. A year ago, voice modes and reasoning model launches were the marquee install events. Now the spike comes when the picture generator gets better.
Appfigures attributes the gap to the visceral, shareable nature of image output. A new reasoning benchmark is abstract; a viral image trend isn't. Nano Banana, the codename for Google's Gemini 2.5 Flash image model, became a social-feed phenomenon in a way that prior Gemini text upgrades did not.
Key facts
- 01Image model releases generate 6.5x more app downloads than standard model updates, per Appfigures.
- 02Google's Nano Banana drove 22+ million Gemini downloads in 28 days, lifting installs more than 4x.
- 03ChatGPT added 12 million incremental installs in 28 days after the GPT-4o image model launched in March 2025.
- 04OpenAI converted that attention into $70 million in gross consumer spending; Nano Banana produced just $181,000.
- 05Meta AI's Vibes added 2.6 million downloads after its September 2025 launch with no meaningful revenue.
Downloads, however, are not revenue. This is where the report turns awkward for Google. Nano Banana drove $181,000 in estimated gross consumer spending in its 28-day window — a rounding error against the 22 million install surge it generated.
“Nano Banana added 22 million Gemini downloads in 28 days but only $181,000 in consumer spending. ChatGPT's GPT-4o image model converted attention to $70 million.”— Jaeden Schafer
OpenAI is the outlier that turned attention into dollars. ChatGPT's GPT-4o image model produced an estimated $70 million in gross consumer spending over its 28-day window, several orders of magnitude above what Google captured from a larger download spike. Meta AI's Vibes, the video feed launched in September 2025, added 2.6 million downloads with no meaningful revenue lift.
The disparity points to a monetization gap rather than a product gap. Google has Gemini bundled inside a sprawling free product surface — Search, Workspace, Android — and converting a viral image moment into a paid Gemini subscription is harder when most users already get the headline feature for free. ChatGPT funnels users toward a $20-per-month upgrade prompt almost immediately.
DeepSeek sits outside the framework entirely. DeepSeek R1 drove 28 million downloads after its January 2025 release, but Appfigures flagged this as a non-comparable event — the moment DeepSeek went from obscure to global story on the strength of its training-cost claims, not an image feature. Curiosity, not capability, was the install driver.
For OpenAI and Google, the read-through is that the next model cycle will weigh image and video heavily, even when the underlying research progress is happening on the reasoning side. App-store growth dictates roadmap priorities, and the chart says pictures.
There is a counterargument worth naming. A 28-day window captures a launch spike, not durable engagement, and Appfigures' own data shows downloads decoupling from spending. If image model releases pull in users who churn before paying, the labs are buying top-of-funnel volume that doesn't compound. Google's Nano Banana numbers fit that profile cleanly.
The strategic question for the consumer AI category is whether image generation is a wedge or a treadmill. Image features get users to install — but the labs that can route those users into paid reasoning, coding, and agentic workflows are the ones that will translate the install spike into the recurring revenue investors are paying current valuations for. On the evidence so far, OpenAI is the only one doing it at scale.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
