Meta enters AI image model race in bid to court advertisers and subscribers
CNBC ● Covered by 2 sources
Meta just launched Muse Image, its own AI picture-generator, built to lure advertisers and paying subscribers. It's Meta's bid to stop leaning on outside AI tools and start making money off its own.
Based on reporting by CNBC — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Meta spent Tuesday introducing Muse Image, an AI model built to generate pictures, and the timing says a lot about where the company thinks its next dollar is coming from. This is the second big release out of Meta Superintelligence Labs, the unit run by Alexandr Wang that already put out the Muse Spark language model back in April as the successor to Llama. Muse Image was apparently known internally as Mango before it got a public name, which is the kind of detail that tells you this thing had been cooking for a while.
For regular people, access is free through the Meta AI app and site, WhatsApp DMs, and Instagram Stories. But free only goes so far — creators and power users who want to generate a lot of images, or unlock certain features, need one of the monthly subscription plans Meta rolled out back in May. Hit your free limit and you either wait for it to reset or pay for a Meta One subscription. That's a fairly blunt monetization lever, and it's clearly intentional.
The advertiser angle is where this gets more interesting. Muse Image is being built into Meta's Advantage Plus ad tools, letting brands tweak elements, swap styles, and spin up variations of their creative without redoing everything from scratch. Meta says advertisers and agencies should start seeing image variants powered by the new model in the coming weeks. Given how much Meta has poured into AI infrastructure, turning that spending into an actual ad-revenue tool is the whole point here — this isn't just a fun toy for Instagram Stories.
Meta isn't first to this particular party. OpenAI and Google both got there earlier, and Google's Nano Banana became a genuine consumer hit when it launched last fall. Meta's own internal benchmarks put Muse Image behind OpenAI's GPT Image 2, but ahead of Nano Banana 2 on editing tasks involving single and multiple images — which is a mixed result Meta seems comfortable publishing anyway. Up to now, Meta has been renting image and video generation muscle from outside shops like Midjourney and Black Forest Labs. The stated plan is to lean less on those partners going forward, and there's a video-generation sibling, Muse Video, already promised for later, with Meta touting strong prompt adherence and visual consistency. Muse Image itself is set to spread further into Facebook, Messenger, and more corners of Instagram and WhatsApp before the year is out.
My take — AI-written commentary, not fact-checked reporting
Nobody should be shocked that Meta's push into AI image generation runs straight through advertisers, not artists — that's the business, and pretending otherwise would be the real surprise. The benchmark humility here is oddly refreshing; admitting you're behind GPT Image 2 while claiming a win over Nano Banana 2 is a small, calculated flex rather than the usual chest-thumping. The bigger story is Meta trying to stop paying rent to Midjourney and Black Forest Labs — owning the pipeline end to end is the actual prize, not whichever model wins this week's leaderboard.
Read more about this at: CNBC