Meta steps up AI push with launch of Muse Wonder models
Meta described the models as its most capable AI models to date, designed to handle complex tasks for personal and professional use.
2 min read
In a blog post on October 12, the US-based tech giant said the launch marked the next stage of its AI development efforts, following a major overhaul that began 14 months ago.
"In that time, we’ve gone from rebuilding our stack from first principles to pushing the frontier," the company said.
Meta described the models as its most capable AI models to date, designed to handle complex tasks for personal and professional use, including software development, research and other work requiring multiple steps.
The models build on the groundwork laid by Muse Spark and are designed to work on complex tasks over extended periods. Meta said they combine two capabilities it calls Active Memory and Active Focus.
Active Memory allows the models to retain important context across conversations and retrieve relevant information when needed. Active Focus enables them to work towards a goal, try alternative approaches when they encounter problems and adapt to new information.
These capabilities are intended to make the models more useful for tasks requiring sustained effort, rather than simply responding to individual prompts.
Meta said the models can handle specialised work in areas such as law, finance and healthcare, alongside complex coding tasks. The company cited examples such as interpreting a lease, understanding financial information and building applications.
It also highlighted the models’ ability to work across software tools and complete tasks involving multiple stages of planning, execution and revision.