Meta has released Muse Spark 1.1, an upgraded artificial intelligence model built for complex coding and autonomous tasks.
The launch signals a shift in how Meta, the owner of Facebook and Instagram, intends to compete in the crowded market for frontier AI systems.
Mark Zuckerberg, the company's chief executive, has poured resources into AI development and accelerated the release schedule for its most advanced models.
Analysts increasingly view Meta as capable of operating as a full frontier laboratory, pointing to its energy supplies, data centres and vast pools of training data.
The most significant change may be commercial rather than technical.
Meta is preparing to sell access to its models through an application programming interface (API), the software bridge that lets outside developers plug the technology into their own products.
That would give the company a potential revenue stream separate from its core advertising business.
Early internal estimates, benchmarked against Anthropic's sales, suggest the API operation could generate more than $1 billion a year.
The size of the opportunity remains uncertain, since it is difficult to forecast how many developer requests Meta will capture in a fast-moving market.
Selling model access would also help Meta spread the cost of training, which can run to billions of dollars, across a wider set of products.
The company could distribute its models through smart glasses, creating a new computing platform with improved voice features.
Enterprise sales have long troubled Meta, which has tried and largely failed to build a business beyond advertising.
Previous efforts included buying a customer relationship management tool and developing AI agents for businesses, with little to show for it.
Some observers remain sceptical that Meta can succeed where costly ventures into virtual reality have drained attention from its profitable advertising arm.
A more modest route would see Meta rent out its graphics processing units to other AI firms, earning short-term income without building enterprise infrastructure.
Rivals offer both cautionary and encouraging precedents.
Amazon and Google turned consumer operations into thriving enterprise arms through Amazon Web Services and Google Cloud.
Microsoft, by contrast, has struggled for years to bridge consumer and enterprise markets.
Meta is expected to keep raising capital expenditure, with some 2027 forecasts ranging from $200 billion to $250 billion.
That spending sits alongside heavy investment from Amazon and Google, underlining an AI infrastructure race that shows no sign of slowing.
Pricing is emerging as the key battleground, with Meta's model potentially undercutting competitors including Anthropic's Fable 5.
Zuckerberg broke a long silence on X to promote the new model, a sign of how fiercely companies are now fighting to shape opinion.
Elon Musk struck a supportive tone on X, praising rival Anthropic in a rare gesture, given that it is thought to be a customer of his rocket company SpaceX.
